Changelog

What changed, and when

This archive is a living dataset. Records are added, classifications are revised, and figures move as a result. Anything that alters what the corpus asserts, how a record is categorised, or how a published number is computed is listed here with its date.

Corrections to individual records are handled separately, through the corrections and right of reply process. If you cited a figure from this archive and need to know whether it has since moved, this page and the dataset downloads are the places to check.

July 29, 2026 records

A woman executed for sorcery, and why the archive will not call her a fraudster

Al Jazeera turned out to be an under-used source, and the first case it produced is one the archive had to be careful with.

Asked whether Al Jazeera had been searched, the answer was: barely — six records cite it out of more than thirteen hundred. Searching it properly produced a case the archive did not hold.
On 12 December 2011 Saudi Arabia beheaded Amina bint Abdul Halim bin Salem Nasser, a woman reported to be in her sixties, in the northern province of Jawf. She had been convicted of practising witchcraft and sorcery. Amnesty International counted it as the kingdom's seventy-third execution that year.
A member of the religious police told a newspaper that she had tricked people into giving her money by claiming she could cure their illnesses, charging up to $800 a session. That description is the pattern this archive documents in hundreds of records, and it would have been easy to file her as another healing fraud.
The archive will not do that, and the reason matters. She was not convicted of fraud. She was executed for practising sorcery — an offence that punishes the supposed possession of occult power rather than any harm done with it, and which treats the reality of that power as settled by the conviction itself. No court is named in the reporting. No trial record exists publicly. No defence is described, no appeal recorded, and the report notes that many of those executed in the kingdom had no lawyer and were not told what proceedings were against them.
An untested characterisation from an unnamed police officer is not an adjudicated finding, and this archive is not going to record a woman the state killed as a convicted fraudster on that basis. What is documented is what is recorded: that Saudi Arabia executed her for sorcery. She is filed with the Gambia and Central African Republic entries, where the state is the authority whose conduct is at issue.
She is named, deliberately. Her name was published by the state that killed her and reported internationally, and a person executed by a government should not be anonymous in the record of that execution.
Three Russian judgments were added alongside: fourteen counts in Chita, a Chelyabinsk robbery, a Stary Oskol theft in which the items taken were the victim's own icon pendant and crucifix, and a Makhachkala case where the defendant advertised herself as a fortune-teller with extrasensory powers and took payments in instalments from four women.

Effect on published figures

1,339 public records. Saudi Arabia gains a second record; 164 institutional events.

July 29, 2026 records

Acting on the consciousness and the will

Fourteen women in eight months, and a Russian district court that describes the mechanism in the language this archive was built to document.

Chita, Zabaykalsky Krai, offences July 2016 to March 2017. Fourteen counts against fourteen women — the largest single-defendant series this archive holds. The method never varies: a woman she did not know, seen in the street, told a curse was upon her, asked to take off her gold and hand it over for the rite.
The phrasing the court uses to describe what she did is worth quoting in full, because it is this archive's entire thesis stated by a Russian district judge: she acted «воздействуя при этом на сознание и волю потерпевшей сведениями об имеющейся порче» — working upon the victim's consciousness and will, by means of the information that a curse was present.
That is not a description of a bad bargain. It says the claim about the curse was the instrument, and the thing it operated on was the victim's will. Fourteen women accepted it in eight months, losing between 7,500 and 89,000 roubles each.
She was on parole when she did it. That release was revoked and she went to a general-regime colony for four years.
Chelyabinsk, offence 2017 — a much smaller case that marks a boundary. A ring and 2,000 roubles taken in the street under the same pretext, and then the victim understood and asked for them back. The defendant kept them anyway, knowing she had been seen, and the court classified the offence as open robbery rather than fraud.
That is now the third case in the archive where the belief fails while the offender is still standing there — Bogorodsk, Novosibirsk, and now this. They are worth holding because they show the limit of the mechanism: it works only while it is believed, and the moment it stops, what remains is an ordinary theft that any court can name.

Effect on published figures

1,336 public records.

July 29, 2026 records

A year with the inspectors locked out

Georgia enters the archive through a UN finding about an Orthodox Church orphanage — and two other priority jurisdictions came back empty.

Ninotsminda St Nino Children's Boarding School is a closed-type orphanage in southern Georgia run by the Georgian Orthodox Church. In June 2024 the UN Committee on the Rights of the Child found that Georgia had violated the Convention by failing to intervene against the frequent physical and psychological abuse of children living there. The complaint was brought on behalf of 57 of them.
The sequence is what makes this a record rather than a mention. Georgia's own Public Defender documented the conditions in 2015 — corporal punishment by prostrations, skipping meals, confinement, forced crawling. The orphanage was granted a care licence the following year. A second report in 2018 found the problems systemic. And from June 2020 the Public Defender was refused entry for almost a year.
It took interim measures from the Committee to get inspectors back through the door. Twenty-seven children were then moved to alternative care; fifteen, including at least one disabled child, were still there in November 2021.
Two children's cases were set out. One, born in 2008, was there from the age of three to thirteen and punished harshly for bedwetting; caregivers would instruct older children to beat her with sticks or hands, and she was made to take psychotropic medication at eleven. The other, born in 2003, faced inadequate food, poor hygiene and restricted movement, and her disabled brother was neglected too.
The record is filed as a finding against the state, not a conviction of anyone. The Committee asked Georgia to investigate and prosecute those responsible; whether that happened is not established here.
A second Georgian case was examined and left out. Archpriest Giorgi Mamaladze is serving nine years for attempted murder, but his trial was held behind closed doors, the prosecution's recordings were never released, a human rights organisation found the proceedings violated his right to a fair trial, and the intended victim was a fellow cleric rather than anyone under his authority. A conviction from a closed trial is not something this archive can rely on.
Armenia and Estonia, the other two jurisdictions ranked as most promising, produced nothing. What Armenia offers instead is a government prosecuting an archbishop, a bishop, twelve priests and the Catholicos amid an open church-state confrontation — political charges that this archive should not treat as neutral adjudication.

Effect on published figures

1,334 public records. Georgia enters the archive; 163 institutional events.

July 29, 2026 records

Two French bishops convicted for not reporting — and a third correctly left out

A jurisdiction audit found France holding three Catholic records against a national commission's estimate of 216,000 victims. The first additions are the bishops convicted for concealment.

Asked whether there must be more public records of Catholic abuse, the honest answer is that this is already the archive's deepest vein — 331 Catholic records, 237 of them convictions, across more than thirty countries. But the audit that followed found the distribution badly uneven, and France is the clearest case: three records, against the CIASE commission's estimate of 216,000 victims of clergy since 1950.
Two convictions now added, and both concern concealment rather than abuse. Pierre Pican, Bishop of Bayeux and Lisieux, was convicted in September 2001 of failing to report the offending of abbé René Bissey, and sentenced to three months suspended. He declined to appeal, so the conviction is final. It was the first conviction of a bishop by the French courts since the Revolution. Bissey himself had been sentenced the previous year to eighteen years.
André Fort, emeritus Bishop of Orléans, was convicted in November 2018 of the same offence and given eight months suspended, over his failure to report a priest who had been chaplain to the Scouts d'Europe. Three of that priest's victims brought the complaint against him in 2016.
A third bishop is deliberately absent, and the reason is the point of this entry. Philippe Barbarin, Archbishop of Lyon, is routinely listed alongside the other two. He was convicted at first instance in 2019 over his failure to report Bernard Preynat — whose own conviction this archive already holds. But the Lyon Court of Appeal acquitted him on 30 January 2020, and the Cour de cassation upheld that in April 2021.
So there is no Barbarin conviction to record. He appears in this archive only in the scope notes of the two bishops who were convicted and of Preynat, always alongside the acquittal, and never as a subject. Listing him with the others would have been an accusation the courts specifically rejected.
Both new records state plainly that the offence is failure to report and not participation in abuse. Fort's appellate position is marked unchecked rather than none-found, because the source does not say whether he appealed — and in this specific category, as Barbarin shows, an unchecked first-instance conviction is exactly the thing that must not be presented as settled.

Effect on published figures

1,333 public records. France rises from three Catholic records to five.

July 29, 2026 records

Ten days of rites, and the smallest theft in the archive

Two Siberian judgments that between them mark the patience and the floor of this pattern, and a military-chaplain search that produced nothing new.

Prokopyevsk, Kemerovo Oblast, offence January 2017 — a new region. The court found the scheme was devised before anyone was approached: she 'planned and developed a plan to steal money from citizens by deception, under the pretext of removing a curse'.
Then she waited. She met a shop assistant at her counter in a children's clothing shop, told her a curse was upon her, and spent ten days conducting what the judgment calls 'sham healing rites' — performances with no purpose but to establish that she was what she claimed. The court treats those ten days as part of the offence, not as preamble.
Only then, in the victim's own flat, came the request: a mink coat, cash, a set of underwear and a bottle of perfume, all said to be needed for the healing, all to be returned the next day. 101,500 roubles. The underwear and the perfume are the detail that matters — a mink coat is what any fraud takes, but those two items are only intelligible as things a rite was said to require.
Novosibirsk, offence July 2017 — the other end of the same range. A woman saw gold earrings in a stranger's ears in the street, asked for them to remove a curse, and took three earrings and 101 roubles: barely more than a bus fare, and the smallest taking this archive holds for the pattern.
Set against the 8,182,000 roubles taken from one victim in Tver, or the four million in jewellery in Gatchina, that is the useful comparison. Nothing about the method differs between them. The claim is not calibrated to what the target has — it takes whatever is in reach, and the sums in this corpus reflect the victims' circumstances rather than any judgement by the offender about what is worth attempting.
Separately, a search for military chaplains abusing people under their pastoral charge produced no new record. The archive already holds four — US Air Force, Army, Navy and Canadian Forces — plus two prison chaplains and the New Zealand Anglican finding naming two school chaplains. Two leads were examined and both correctly excluded: an American chaplain convicted over an affair and a threat, where the woman was not under his pastoral charge, and an Austrian military priest accused in 2017 of harassing conscripts, where no charge, trial or outcome was ever located. An allegation without a disposition is not a record here.

Effect on published figures

1,330 public records. Kemerovo Oblast enters the archive.

July 29, 2026 Correction

Two guru cases were being counted twice

An observation that guru cases looked oddly split between Hinduism and new religious movements turned out to be a double-count, now fixed without losing any evidence.

Two Indian cases were in this archive twice, filed under different traditions. Gurmeet Ram Rahim Singh of Dera Sacha Sauda appeared both as a legacy entry under Hinduism and as a full record under new religious movements. Nithyananda appeared the same way. So both traditions were inflated by the same events, and the guru category looked 'split' when part of what was happening was one set of facts counted twice.
Every per-tradition figure this archive published was affected. Hinduism now stands at 29 records and new religious movements at 69; core records fall from 1,110 to 1,108, which is exactly the two duplicates and nothing else.
A third case looked like the same problem and was not. Asaram Bapu has a legacy entry for his 2018 Jodhpur conviction and a full record for a separate 2023 Ahmedabad conviction by a different complainant. Those are two prosecutions, not one duplicated. Withdrawing the legacy entry would have erased a conviction; it stays.
The fix required a mechanism this archive did not have. The original 71 legacy records are protected and cannot be edited, so a duplicate could not simply be removed. There is now a withdrawal that names its replacement, so the change is auditable rather than a silent deletion, and the build refuses to run if the named replacement does not exist.
Withdrawal never destroys evidence. Every source the withdrawn entry cited transfers to the surviving record — which is why the archive's source count rose from 2,378 to 2,381 while the record count fell. The build already enforced that every legacy citation must appear byte-for-byte in the published data, and that guarantee is the reason a duplicate can be withdrawn safely at all.
Two tests failed when this landed, and they were right to. One asserted that all 71 legacy entries resolve to a published record; the other pinned the count of cases still resting partly on Wikipedia. Both were rewritten to state the new guarantee rather than to accommodate the change: the first now checks that a withdrawn entry is gone AND that its sources appear on the named survivor, which is a stronger promise than the one it replaced.

Effect on published figures

Core records 1,110 to 1,108. Hinduism 29, new religious movements 69. Sources 2,381. A duplicate withdrawal mechanism now exists, with the withdrawal named, asserted and tested.

July 29, 2026 records

Central African Republic: the state itself convicts the accused

A Sahel coverage audit found the region almost dark, and the first Central African Republic record is the archive's criterion running in reverse.

Prompted to look at the Sahel, the audit confirmed it: coastal West Africa is covered — Nigeria 18 records, Ghana 6, Senegal 6 — but the core Sahel is nearly empty. Mali 2, Burkina Faso 3, Chad 1, Mauritania 1, and Niger and the Central African Republic at zero. Of those, only three are actual convictions; the Chad and Mauritania entries are UN treaty-body findings.
The Central African Republic now has its first record, and it is the inverse of what this archive usually documents. Articles 149 and 150 of the Central African Penal Code make 'charlatanism or witchcraft practices likely to disturb public order or harm people or property' a crime. Avocats Sans Frontières, which provides legal aid to the accused, reports that at Bimbo women's prison half the women held are convicted of alleged witchcraft offences.
The mechanism the study describes is the mirror image of the one traced elsewhere in this corpus. Where a healer's claim makes a demand unrefusable, here a court's conviction makes an accusation permanent. Faced with the impossibility of material proof, most judicial actors treat the accused person's confession as conclusive — 'regardless of the motives of the accused in making the confession', which the study notes is often given for social appeasement or self-protection.
And conviction is not the end of the harm: 'By prosecuting a person for CWP, the court attests to the reality of her or his witchcraft', so the person remains exposed to further prosecution and further violence after release. Acquittal is no safer — a population distrustful of the courts 'may seek justice for themselves'.
Accusations fall on isolated and elderly women. That is the same victim profile as the curse-removal frauds this archive has been reading all week in Russia, reached from the opposite direction: there the elderly woman is defrauded by someone claiming power over curses, here she is imprisoned for supposedly having it.
Filed as accusation-violence, an institutional event, with the convicted party marked 'other' — the convictions run against the accused, not against any spiritual authority, and a reader filtering this corpus for convicted practitioners must not pick it up. The source is an NGO field study, not a judicial document, and the record says so. A much broader country-wide proportion circulates in secondary coverage and is deliberately not reproduced; the figure recorded is the one ASF states for one prison.

Effect on published figures

The Central African Republic enters the archive. Institutional events rise to 162.

July 29, 2026 records

The clothes off the mannequins, and a coincidence the archive can't resolve

A Tatarstan judgment about two market traders, and an honest limit on what anonymised judgments let this archive know.

Naberezhnye Chelny, offences March 2016 and April 2017. Both victims were traders working their own pitches — a flower kiosk on the ground floor of a services building, then a pavilion at a wholesale market. Somewhere a trader cannot walk away, and where a stranger stopping to talk is the ordinary business of the day.
The first handed over 1,000 roubles, a gold signet ring and gold earrings: 31,000 roubles. The second handed over a gold chain — and, because the rite was said to require them, the velour dressing gown and the jumper hanging on her mannequins. Her own stock, off the display.
That is the detail worth keeping. Once a rite defines what it needs, there is no natural point at which the person being asked can say the object is the wrong kind of object. Total 42,500 roubles; three months in a penal colony; compensation paid to one victim in full and one in part before sentence.
A limit worth stating plainly. The archive already held a Naberezhnye Chelny case from the same court whose offence window overlaps this one, and both defendants are anonymised as ФИО2. On the face of the documents they are separate prosecutions — different case numbers, judgments thirteen months apart, one victim there against two here, a continuing course of requests there against two discrete encounters here, and neither judgment mentions the other.
But both women are masked, so the archive cannot exclude that they are the same person. Both records now carry that disclosure, so neither is read as establishing a separate offender. Russian anonymisation makes some questions unanswerable, and saying so is more useful than a confident count.

Effect on published figures

Russia rises to seventy-eight records. Eighty-four judgments on this practice have now been read across Russia and Ukraine, of which eighty-one qualified.

July 29, 2026 records

Holy water, fetched from the kitchen while the bedroom was emptied

A Tver judgment in which the rite does two jobs at once: it opens the door, then it empties the room.

Tver, Proletarsky District Court, offences January 2017. Four counts — a large-scale theft with unlawful entry, two frauds, and an attempt that failed for reasons outside the defendant's control. Five years in a general-regime colony, one of the longest real terms this archive holds for the pattern.
The court is explicit that the curse claim was made to get inside: she approached the householder «для облегчения... проникновения в указанный дом», told her a curse lay on her and that it could be lifted, and the woman invited her in.
Then came the second use. She asked the householder to bring her holy water for the removal. The judgment records what happened next in a single clause — on receiving the glass she «быстрым шагом проследовала в комнату», walked quickly into the bedroom — and took 273,000 roubles of jewellery.
Neither step needed force or stealth. The rite supplied a reason for the door to open and a reason for the owner to leave the room, and the borrowing of an Orthodox object lent the request a legitimacy a purely folk demand would not have carried. Immediately afterwards, still in the same house, she ran a second deception on the same woman.
Two further victims lost 5,000 and 24,500 roubles; all three were awarded compensation.
As with the Tomsk record added earlier today, no parish, priest or religious body is implicated. Holy water appears only as an object the defendant asked the victim to fetch, and the record says so — a corpus that filed this under an Orthodox heading would be making an accusation the judgment does not support.

Effect on published figures

Russia rises to seventy-seven records. Eighty-three judgments on this practice have now been read across Russia and Ukraine, of which eighty qualified.

July 29, 2026 records

A court names the mechanism: 'under her psychological influence'

One judgment from rural Omsk that states this archive's inclusion test more plainly than any other record it holds.

Moskalensky District, Omsk Oblast, offences April 2017. Two elderly women in their own homes, three days apart. The defendant was lawfully inside on both occasions — no trespass, no forced entry, nothing to raise alarm.
What the court found is the part worth quoting. Not merely that the victims were deceived about a service, but that each acted «под её психическим воздействием, вызванным разговором на темы о здоровье» — under her psychological influence, produced by conversation about her own health and the health of her relatives. The deception is described as being about «целебной силы совершаемых ею действий», the curative power of the acts she performed, for removing a curse and treating the family.
That is this archive's criterion in a court's own words: an authority claimed over health and misfortune, and a victim who complies because of it rather than because she has weighed a price. The property was asked for on a promise of return, so no such moment arose.
The first victim gave up the gold earrings she was wearing, valued at 20,000 roubles. The court recorded why that mattered: her pension is 10,000 roubles a month, she has no home of her own, and she lives with her daughter. The earrings were worth two months of everything she has. The second victim gave 2,800 roubles and a mobile telephone.
Everything was recovered — the earrings handed in voluntarily by the defendant's son, the cash returned at a police station — and neither woman brought a civil claim. Sixteen months, suspended.

Effect on published figures

Russia rises to seventy-six records. Eighty-two judgments on this practice have now been read across Russia and Ukraine, of which seventy-nine qualified.

July 29, 2026 records

Four courts in one city, and a cross taken from a stranger's neck

Two more judgments from the register, and a pattern in the archive's own geography worth naming.

Krasnoyarsk, judgment October 2017. A woman struck up an acquaintance with a stranger and asked her for a chicken egg, her gold earrings and 2,000 roubles, promising to return them shortly. 12,000 roubles; repaid in full; eighteen months, suspended.
The egg is the detail that recurs. It appears across this archive's Russian judgments — broken into a bag in Tver, produced for the rite in Karpinsk, requested here — and its function is consistent: asking for an egg alongside the gold makes both read as materials the rite requires rather than as things being taken. Combined with a promise to return the valuables afterwards, it removes the moment at which a person weighs a price. Nothing is being bought, so nothing has to be judged worth the cost.
That is the fourth Krasnoyarsk case in this archive, and the fourth distinct district court in that one city — Oktyabrsky, Kirovsky, Leninsky and now Sverdlovsky. Four separate benches in a single city have convicted for this pattern. The archive does not draw conclusions about prevalence from its own coverage, since coverage follows what happens to be published; but four courts is worth recording plainly.
Chita, Zabaykalsky Krai, offence July 2017 — a new region, on the eastern edge of the country. The most compressed version of this offence the archive holds. No diagnosis was delivered in advance and no relationship was built: the court found the intent arose the moment the defendant saw a gold chain and cross on a stranger's neck outside a cinema.
She asked for them 'for the performance of the rite', knowing she would not return them, and the victim took them off and handed them over. What was taken is worth pausing on — a cross the woman wore around her neck, an object of devotion rather than ornament. The frame that made it surrenderable was the same one that made surrendering it feel like taking part.
13,000 roubles. Her earlier suspended sentence was revoked and she went to a penal colony for three years.

Effect on published figures

Russia rises to seventy-five records and Zabaykalsky Krai enters the archive. Eighty-one judgments on this practice have now been read across Russia and Ukraine, of which seventy-eight qualified.

July 29, 2026 records

Eight front doors opened to a request for alms

Two judgments opening a new page of the register, one of them the widest single series the archive holds from one defendant.

Krasnoyarsk, judgment December 2017. Eight counts against eight households — seven frauds and one open theft — and the same sequence each time. The defendant was at the door asking for alms. Once in conversation, she told the occupant a curse lay on her valuables and asked for the gold or the cash so it could be charmed, on the stated condition that the charming happen outside the flat.
The alms request is what made eight separate front doors open. Someone asking for charity has a reason to be there that householders recognise, and refusing feels ungenerous before any claim has been made. The curse then converted that encounter into a transaction over the household's own property, and the requirement that the rite happen outside gave a reason — inside the rite's own terms — for the valuables to leave the owner's sight.
Recorded losses: 7,100, 86,000, 12,000, 27,000, 20,400, 27,000, 38,000 and 19,500 roubles. Two years in a colony-settlement. The judgment redacts every date as 00.00.0000, so this record states none.
That is the third Krasnoyarsk case in the archive and the third distinct court in that one city.
Tikhvin, Leningrad Oblast, offence May 2017. Two women on a footpath. The first told the victim she was under a curse and that her companion had the power to heal — and only then did the companion speak, confirming it on her own account.
That division of labour is the mechanism. A stranger asserting her own healing power invites doubt; a stranger vouching for someone else's presents as a witness rather than an interested party. The rite was then defined as a charm laid on the victim's cash and gold, which is what required her to produce everything rather than pay a fee — the property was the thing being treated. 108,000 roubles. Three years six months in a penal colony after aggregation with an earlier sentence.
The second defendant's individual term sits in a passage of the judgment that is not fully legible. The archive states the first defendant's sentence and leaves the second's blank rather than estimating it.

Effect on published figures

Russia rises to seventy-three records. Seventy-nine judgments on this practice have now been read across Russia and Ukraine, of which seventy-six qualified.

July 29, 2026 records

Two Orders of the Patriotic War, and a case the archive left out

Two judgments closing this page of the register, and a third that documents a real fraud the archive still cannot count.

Tambov, offence from December 2017. A woman working in a curtain kiosk was offered help removing a curse, across successive visits under what the court found was a single continuing intent. What she handed over was 179,347 roubles of gold — including both wedding rings, hers and her husband's — and then, on a later visit, her family's Soviet state decorations: two Orders of the Patriotic War, a medal for the 30th anniversary of victory, and three armed forces medals.
Those are not objects a person parts with for a service she is weighing up. That they were surrendered at all is the measure of what the claim was doing — they were required for the removal, so giving them up was a step in the remedy. A separate charge under article 324 covers the unlawful acquisition of the decorations; the medals were returned to her, and a television returned in partial compensation. One year in a colony-settlement.
Bogorodsk, Nizhny Novgorod Oblast, offence 2018. The only judgment in this archive where the belief fails while the offender is still in the room. The pretext worked as it always does — the occupant was asked to fetch her own jacket, earrings and 120 roubles because the ritual required them, and she did. Then she realised, and ran out into the street after her, demanding them back.
The court found the defendant knew she had been seen and fled anyway, which converted the offence from fraud into open robbery under article 161. The record is kept because it shows the same mechanism at the point of collapse: the belief secured the property, and the moment it lapsed there was nothing left but a woman running down a street with someone else's coat. The judgment's stated motive was money for alcohol. Every date and sentence figure in that judgment is redacted, so this record states none of them.
One judgment was read and left out, and it is worth saying why. A police investigator in Oryol was convicted of abuse of office for falsifying a report. The case she was investigating was a curse-removal fraud in which a girl and her mother lost gold — a real offence, opened in May 2016, described in court as causing significant public resonance.
But that fraud was recorded against an unidentified person and no one was ever charged with it. The only conviction in the judgment is of the officer, for what she did to the file. The archive holds adjudicated outcomes, not open investigations, and it does not count a crime because a court mentioned it in passing. The girl and her mother are in the record of this changelog and nowhere else, which is the honest place for them.

Effect on published figures

Russia rises to seventy-one records. Seventy-seven judgments on this practice have now been read across Russia and Ukraine, of which seventy-four qualified.

July 29, 2026 records

A live chicken, a white sheet, and a rite that was supposed to be the Church's

Two judgments from Tatarstan and Tomsk, the second invoking an authority the rest of this set does not.

Zainsk, Tatarstan, judgment September 2018. Two women, three victims, 292,397 roubles. The rite is set out in the judgment with unusual precision: a white sheet, a live chicken, thread, water and sugar.
The specificity matters because the court treats those objects as the instrument of persuasion — the manipulations are what convinced the couple the curse was real. And the demand for a live chicken did something else: it required the victim to prepare in advance and to admit the women to her home, so by the time anything was asked for, she had already invested in the procedure.
The curse was said to be on the husband, not on the woman approached at the market. That pattern recurs across this archive — in Ufa the curse was on the child's mother, here on the wife's husband. Placing the affliction on someone the victim loves makes compliance read as protection rather than as a purchase.
On sentence: each defendant was fined, then released from paying entirely under article 72(5) in view of time already spent in pre-trial detention, and freed in the courtroom. Neither paid a fine nor served further custody. The record says so, because printing the fine figures alone would describe a penalty that was never enforced. Jewellery and 11,000 roubles held as exhibits went back to the victims.
Tomsk, offence April 2018 — a new oblast, and a different kind of claim. The defendant did not say she could lift a curse by her own power. She said she could arrange a church cleansing rite, asked for 1,000 roubles to buy church candles for it, then asked for the victim's gold wedding ring as the rite required, and left saying she was going to the church.
That framing does work a folk claim cannot. The money and the ring appear to be going to a church rather than to the person asking, which makes the handover feel less like payment and harder to refuse without seeming to refuse the Church itself. The departure was supplied by the same frame.
No parish, priest or religious body is implicated in that judgment, and the record says so explicitly. It stays classified with the divination-fraud pattern rather than under any Christian tradition, because what the court found was a woman invoking the Church as a pretext — not the Church doing anything.

Effect on published figures

Russia rises to sixty-nine records and Tomsk Oblast enters the archive. Seventy-four judgments on this practice have now been read across Russia and Ukraine, of which seventy-two qualified.

July 29, 2026 records

Two prosecutions, one offence: a merge rather than a second record

A companion judgment that would have inflated the count, and two additions from Tver and Bryansk.

The Crimean record added earlier today — a woman who entered a house under a curse-removal pretext while an accomplice occupied a bystander with an offer to buy feathers and watches — has a companion judgment. It prosecutes the other participant, the one who ran the distraction, and it is the same offence: same day, 11 June 2018, same victim, same 52,468.62 roubles.
It has been merged into the existing record as a second proceeding rather than added as a new one. Two prosecutions of one episode are not two episodes, and counting them separately would double this archive's figure for that victim and for this pattern. The record now carries both judgments, and its scope note says why there is one entry and not two.
Both participants have now been convicted. A third — the driver in one judgment, and in the other a woman whose case was severed because she was being sought — is described differently in the two texts and adjudicated in neither. The archive makes no finding about that person.
Tver, Proletarsky District Court, offence June 2018. The approach was commercial: the defendant came into a clothing alteration shop as a customer, ordering a leather jacket and skirt, and only then offered the proprietor a card reading — asking her to lock the shop door from the inside so they would not be disturbed. The victim locked her own premises.
The diagnosis arrived inside that locked room, and it was aimed at her family rather than at her: an evil eye and a curse on them. Rites began at once, her gold ring went first, and the defendant said she would return the next day for a second ritual. She did. 151,000 roubles in total, from a woman the court noted earned 20,000 a month. Repaid in full before sentence; eighteen months, suspended.
That is the third Tver case in the archive and the third distinct court in that oblast, none of them double-counted.
Bryansk, offence July 2018. Ten minutes in the street and 10,001 roubles, taken on a pretext combining curse removal with improving the victim's finances — the same claim reaching for both a spiritual worry and a material one. Sixteen months in a colony-settlement: real custody, for a sum smaller than most suspended sentences in this set.
That case was decided on an admission without a trial, so the judgment holds the charge and nothing more — no account of the encounter, no words attributed to anyone. The record says so rather than filling the gap with the pattern seen twenty times elsewhere.

Effect on published figures

Russia rises to sixty-seven records. Seventy-two judgments on this practice have now been read across Russia and Ukraine, of which seventy qualified — one of those merged into an existing record rather than counted separately.

July 29, 2026 records

Lying on an empty bundle of newspaper, told not to get up for an hour

Three judgments closing this page of the register, each showing the rite doing a different kind of work.

Kichmengsky Gorodok, Vologda Oblast, offence January 2018. The most complete example the archive holds of a rite built so the victim guards the theft herself.
The money was required for the removal, so it had to be produced — 74,500 roubles. It was wrapped in newspaper in front of her, so she saw where it went. She was persuaded to lie down on the sofa and the bundle was placed under her back, which is about as secure as a person can believe her savings to be. The notes had already been drawn out. Then she was told, as part of the procedure, not to get up for an hour.
Every step is intelligible only from inside the belief, and together they removed every chance she had to check. Eighteen months in a general-regime colony — real custody, which is uncommon in this set.
Chelyabinsk, offences December 2017. Two women agreed in advance to win the trust of elderly people and enter their homes, with roles fixed: one to hold the occupant's attention and block her view, the other to leave the room and search it. Here the curse pretext was not used to extract anything — it was used to find out where the money was kept, after which the victim was walked into the kitchen. 97,000 and 150,000 roubles from two victims. Three years, suspended.
Naberezhnye Chelny, Tatarstan, offences November 2016 to April 2017. Not one encounter but five months of them. A ring in early November, earrings at the end of it, then further gold and cash — each asked for separately, each on the same basis of removing a curse. 154,900 roubles in total.
That structure works because the belief does not expire when an item is handed over. The removal is always still in progress, so a person who has already given has a reason to give again rather than to conclude she has been robbed. The court found the defendant knew from the very first request that she would return nothing.
One boundary worth stating: the Vologda defendant was convicted on two counts, but only the second involved a curse. The first, eight months earlier, was an ordinary deception in which she posed as a social worker checking banknotes. The record says so, so that an eighteen-month sentence is not read as punishment for two spiritual frauds.

Effect on published figures

Russia rises to sixty-five records. Sixty-nine judgments on this practice have now been read across Russia and Ukraine, of which sixty-seven qualified.

July 29, 2026 records

A thirteen-year-old alone at home, and a curse said to be on her mother

Three more judgments from the Russian register, one of which turns this pattern on a child.

Ufa, Bashkortostan, offence 2018. A woman rang the doorbell of a flat where she knew no one. The complainant was at work; the door was opened by her thirteen-year-old daughter, alone in the flat.
The construction the court found is exact, and it is why this record matters. The curse was said to be on the mother, not on the girl. So fetching her mother's jewellery box and handing it to a stranger presented itself to a child as helping her mother — not as a transaction she had no standing to make. The judgment records that the defendant exploited the child's trustfulness. No threat was used and no price was discussed.
356,840 roubles of jewellery, itemised by weight and standard down to 26.67 grams of scrap gold. Two years six months, aggregated with an earlier Voronezh sentence to three years six months in a colony. The full sum was awarded to the mother.
The published judgment's anonymisation fails for the child: her given name is visible in two passages. It is not reproduced here, and the splice asserts its absence before the record can be written. She appears only as the complainant's thirteen-year-old daughter.
Ishimbay, Bashkortostan, offence October 2018. A knock at a stranger's door, and the valuables asked for under a curse-removal pretext 'on condition of their further return' — a promise the court found was never intended to be kept. That framing removes the moment where a person weighs a price against a benefit: nothing was being bought, so nothing had to be judged worth the cost. 17,088 roubles in cash and gold, itemised to the gram. Eighteen months, suspended.
Shushary, Saint Petersburg, offence September 2018 — a new city for the archive. Gold wedding ring, a ring set with three diamonds, a chain and 2,500 roubles in cash, handed over in a street encounter lasting about twenty minutes and swapped for novelty 'Bank of Jokes' banknotes and three coins.
That is the third judgment in this archive to record those specific novelty notes used for the switch — the other two from Rostov Oblast, a thousand kilometres away. It suggests a prop in circulation for this method rather than anyone's improvisation.

Effect on published figures

Russia rises to sixty-two records and Saint Petersburg enters the archive. Sixty-six judgments on this practice have now been read across Russia and Ukraine, of which sixty-four qualified.

July 29, 2026 records

Five million roubles in Stavropol, and an instruction that made the victim hold the door open

Three judgments opening a new page of the register, including the largest scheme total the archive holds.

Stavropol, judgment May 2019. A single continuing scheme — the court's phrase is «единый преступный умысел» — that took 5,223,157 roubles from more than a dozen people. That is the largest total the archive holds for this practice, though not the largest loss by one victim, which remains the 8,182,000 roubles recorded in the Tver judgment added this morning.
The condition the court sets out is the mechanism in one line: the jewellery and cash were to be handed over «без права на их дальнейшее распоряжение» — the owner surrendering any right to deal with her own property — for as long as the removal required. That turns emptying a household into a term of the rite rather than a price for it.
The sentence was two years. The judgment sets out why: no previous convictions, voluntary surrender, active help tracing the property, three dependent children, a farm to run. The archive records that reasoning because a two-year term against five million roubles is otherwise unintelligible, and this project does not print sentences without the findings that produced them.
This is a second and separate case from the same Stavropol court — distinct from the 1,364,000-rouble judgment added yesterday. Establishing that took comparing case numbers, UIDs, years, prior-conviction histories and whether a co-participant was involved. Different women.
Kandalaksha, Murmansk Oblast, judgment March 2019 — the Arctic north, new to the archive. An elderly woman was told a curse had been placed on her and gave up the gold chain and wedding ring she was wearing. The defendant set a glass of water covered with a bag on the table and told her to drink it after she had gone, and the curse would leave.
The court found the victim let her walk out unhindered because she believed her jewellery was in that glass. The rite did not merely hide the theft — it gave the victim her own reason to hold the door open. That judgment also names the defendant in full, which is unusual for a Russian first-instance decision; the archive does not reproduce it.
Novocherkassk, offence January 2019. A second and separate case from that city, distinct from the September judgment added earlier today. Entry to the flat was obtained on the pretext of help with a difficult life situation; the victim's emerald earrings and gold ring went into a glass of water for the rite and were taken while she was kept talking. 38,000 roubles, a year in a colony-settlement.

Effect on published figures

Russia rises to fifty-nine records and Murmansk Oblast enters the archive. Sixty-three judgments on this practice have now been read across Russia and Ukraine, of which sixty-one qualified.

July 29, 2026 records

Six years in Karelia, and a note on what the archive refuses to repeat

Four judgments completing this page of the register, including the longest sentence in the archive's Russian set.

Petrozavodsk, Karelia, judgment July 2019. Seven victims over two years, 1,067,000 roubles, six years in a penal colony — the longest term the archive holds for this practice anywhere in Russia. The offer was healing by 'methods of non-traditional medicine': removal of curses, sessions of incantation, cleansing from the evil eye.
In every one of the seven counts the court found the same selection criterion — each woman was approached while her family was in serious difficulty, and the treatment was offered as the answer to that difficulty. The sums escalated as households were emptied, moving from cash to jewellery to a set of bed linen worth 950 roubles.
That judgment also states that the defendant invoked her own ethnic background as evidence of inherent healing ability. The archive records that as a finding about the method this individual used — a claim she made about herself — and does not reproduce the term the court used. The pattern in this corpus spans defendants of many backgrounds in a dozen countries, and repeating the label would invite an inference the evidence does not support. A check across all 1,305 records confirms no such attribution appears anywhere in the archive.
Krasnogorsk, Moscow Oblast, judgment May 2019. A woman was persuaded that the jewellery in her flat carried a curse and had to be cleansed; she gathered sixteen pieces and handed them over, 337,351 roubles' worth. When she later asked for them back, she was told they had been flushed into the drain because the negative energy could not be removed from them, and she was told the same thing every time she asked again.
That is the same claim performing two jobs: it justified taking the jewellery, and then it explained the jewellery's absence. A victim who accepted the first had no ground to reject the second.
Bryansk, Sovetsky District Court, judgment July 2019 — a second and separate Bryansk case, distinct from the children's-hospital prosecution added earlier today. Gold rings and earrings handed over for a rite, palmed into the defendant's pocket while she made passes with her hands, and then a 200-rouble fee requested for the service. The victim paid it, believing a rite had been performed.
Kazan, Tatarstan, judgment June 2019. 33,000 roubles in cash, gold and a tablet computer. This one is recorded thinly on purpose: it was decided in the special procedure on an admission, so the judgment contains the charge and nothing else — no account of the encounter, no words attributed to anyone, no description of a rite. The archive states that limitation rather than filling the gap.

Effect on published figures

Russia rises to fifty-six records and the Republic of Karelia enters the archive. Sixty judgments on this practice have now been read across Russia and Ukraine, of which fifty-eight qualified.

July 29, 2026 records

A script written before the victim was chosen, and a record where the archive flags its own limit

Three judgments from Kostroma, Krasnoyarsk and Dagestan — one of which sits at the edge of what this archive can properly claim.

Krasnoyarsk, judgment August 2019. The clearest account yet of this fraud being planned rather than improvised. The court sets out a «заранее подготовленная легенда» — a script agreed before the approach — including who would say what and when.
The victim was selected first: the participants judged that her age made her 'trusting, susceptible and subject to psychological influence'. The curse was to be described as dangerous to life and health. A second woman would arrive at the right moment to confirm that the first was a healer who had cured her own son of cerebral palsy by removing a curse.
The ritual required the victim's money, an egg and her photograph, all to be 'cleansed by prayer'. The money was to be brought in a black sock, swapped for an identical sock of cut newspaper, and the victim told to go home and not open it for three days. All of that is the plan as the court found it was agreed — and all of it happened. 15,000 roubles. Her earlier suspended sentence was revoked; three years in a colony.
Kostroma, judgment August 2019. A woman asked a stranger for a drink of water and followed her inside — which the court held was unlawful entry into a dwelling. She performed divination over the water, then said the money in the house was cursed. The victim fetched 125,000 roubles for the cleansing; 60,000 was concealed and taken, including notes spotted under a box on the sideboard. Charged as aggravated theft, not fraud. Repaid in full; two years suspended.
Dagestan, judgment July 2019 — and this one comes with a qualification the archive is stating rather than burying. Three people drove to a village; two gained a householder's trust and entered 'under the pretext of removing a curse', taking 128,000 roubles from an upstairs bedroom. The third watched the street and drove them away, taking 30,000 roubles as his share.
He is the only one tried in this judgment. The two who actually made the curse claim are not adjudicated in it, and nothing is said about their prosecution. So the record is filed with its convicted party marked 'other' rather than 'spiritual authority', and the scope note says plainly that a reader should not treat it as an adjudication of anyone who exercised a spiritual claim. It is included because the curse pretext is the court's own finding as to how entry was obtained — but the archive would be overstating itself to count this man among convicted healers.
Dagestan is new to the archive, as Chechnya was earlier today.

Effect on published figures

Russia rises to fifty-two records and the Republic of Dagestan enters the archive. Fifty-six judgments on this practice have now been read across Russia and Ukraine, of which fifty-four qualified.

July 29, 2026 records

A widow's wedding rings, four million roubles of jewellery, and a stranger who vouches

Three judgments completing this page of the Russian register, including the largest jewellery loss the archive holds.

Veliky Novgorod, offence April 2019. A cashier was approached at her till. The court found the defendant established in conversation that her husband had died and that she was a trusting person, and only then told her a curse was upon her which she would remove — adding that she would help her bear the loss of her husband.
The rite was built around that loss. The victim was made to write her own name and her dead husband's on a slip of paper, which was pressed to her body. She was then asked whose the two wedding rings on her chain were; told one had been her husband's, the defendant directed both into a glass of water, whispered and blew over it, and lifted them out with scissors. 53,500 roubles. Two years, aggregated with an earlier sentence to four years in a penal colony.
Gatchina, judgment September 2019. The largest jewellery loss in the archive: 3,973,389 roubles, itemised in the judgment piece by piece — emeralds, sapphires, diamonds, topaz, rubies, tourmalines, chrysoprase — taken from a single victim in one transaction.
What made that possible was the framing. The defendant was a personal acquaintance who knew the victim had family problems, and the rite she proposed was a cleansing of the jewellery itself. So the collection had to be physically present to be treated, and handing it over was a step in the remedy rather than its price. Two further sums entrusted to her, 100,000 and 160,000 roubles, were converted afterwards. Five years — suspended in full. Compensation of 4,062,627 roubles.
Rostov-on-Don, judgment August 2019. The corroborator technique again, and here the archive can see exactly what the corroboration was for. A woman was told in the street that a curse was on her; a second stranger arrived moments later and said the first had cured her own mother and had taken no money for it.
That sentence is engineered to defeat the one objection a person would naturally raise — that someone offering healing in the street wants money. It worked: 243,000 roubles went into the woman's handbag wrapped in a bag, an identical empty bag came back, and the victim discovered the loss only on reaching home at five o'clock.
The conviction in that case is of the woman who vouched, not the woman who claimed the power: the prosecution of the latter was terminated after she reconciled with the victim. The archive records the conviction it has and makes no finding about the other.

Effect on published figures

Russia rises to forty-nine records. Fifty-three judgments on this practice have now been read across Russia and Ukraine, of which fifty-one qualified.

July 29, 2026 records

A threat to place a curse, a knotted handkerchief, and an accomplice who arrives on cue

Four judgments from Chechnya, Ulyanovsk, Leningrad Oblast and Rostov Oblast, and one more case the archive turned down.

Ulyanovsk, offence January 2019. The first record in this set where the mechanism runs in both directions. A man registered on VKontakte under a false nickname, offered a woman a divination rite, then told her a curse had been placed on her and that removing it would cost 1,000 roubles. The price rose through the evening — he said he lacked 'material' to install protection, then that the curse was very strong and 4,000 was needed rather than 3,000.
When she said she had no money left, he told her the consequences would fall on her or her parents, and that unless she sent 2,000 more he would place a curse on her. He admitted at the investigation that he said this to frighten her into paying faster. Because the exchange was in writing and payment by bank transfer, the escalation survives step by step. 10,800 roubles, 420 hours of compulsory labour, full compensation ordered.
Gudermes, Chechnya, judgment December 2019 — a new republic for the archive. Two acquaintances mentioned in a grocery shop how much they had saved; each was told a curse lay on that money and that special rites could lift it. The court found the defendant knew she had no such ability. The framing here is tighter than usual: the money was handed over 'for the duration of the rite', with a date for its return. 50,000 and 127,000 roubles. Two years, suspended.
Tosno, Leningrad Oblast, judgment September 2019. A woman stopped in the street and walked to a sports ground to have a curse removed, where she was asked to take off her jewellery for the rite. It was tied into a handkerchief — the knot described to her as a condition of the removal — and handed to her to hold; then untied in front of her so she could see the jewellery was still there, and tied again. 46,000 roubles. Two years, with a further year's restriction of liberty.
Novocherkassk, Rostov Oblast, judgment September 2019. The corroborator technique recorded in full. A woman was told in her own yard that a curse was on her; a second person then arrived on cue and thanked the first, in the victim's hearing, for a healing already given, adding that she had a healing gift. The court found the victim complied 'genuinely fearing for her health'.
She gathered 101,000 roubles and her gold. During the staged rite the valuables went into a New Year gift bag while the accomplice, in her presence, packed novelty 'Bank of Jokes' banknotes into an identical one. 129,000 roubles awarded in compensation; one year suspended. The second woman's case was severed because she is being sought, and the archive makes no finding about her.
One judgment was read and rejected. A woman in Vladimir was convicted of witness intimidation for pressuring a witness during a confrontation at a police station. The underlying case was a 21,000-rouble curse-removal fraud, but that fraud was not adjudicated in this judgment — only the intimidation was. Recording it would mean presenting a charge as an outcome.

Effect on published figures

Russia rises to forty-six records and Chechnya enters the archive as a new republic. Fifty judgments on this practice have now been read across Russia and Ukraine, of which forty-eight qualified.

July 29, 2026 records

A children's hospital, a Crimean courtroom, and a case the archive turned down

Four more Russian judgments, one rejection worth explaining, and a correction to a record published this morning.

Bryansk, December 2019. Two women approached inside a children's hospital on separate days and persuaded to hand over their jewellery so a curse could be lifted — a wedding ring, earrings, chains, five gold rings between them, 38,016 roubles in all. The court found she took the items 'with no intention of returning them'. Eighteen months in a colony-settlement.
The setting is the point. Elsewhere in this pattern the victim is sent home or to a bank; here what was taken was what each woman had on her, in a building people attend because a child is ill.
Veliky Novgorod, offence April 2019. A woman stopped in the street and told she had a bad aura, that she would not live until Easter, and that if she refused the rite she and her relatives would die. She was led to a spot behind railway tracks — the judgment records the coordinates — where a strand of her hair was passed from hand to hand with the words 'As the stream flows, so the curse will go' and thrown into a ditch. 30,500 roubles, repaid; 360 hours of compulsory labour, no custody.
Sudak, offence June 2018. Two women agreed roles in a car: one would occupy a bystander in the yard with an offer to buy feathers and watches, the other would go inside. The curse pretext was not used to extract a payment here but to make the householder reveal where her savings were kept; the money was taken once she left the room. Only one of the three counts involved a curse, and the record says so rather than letting a two-year term read as punishment for three such offences.
That case is filed under Crimea and attributed to neither Russia nor Ukraine in the country totals. The court is a court of the Russian Federation and the conviction is recorded as such; the archive does not treat the status of the territory as settled by the existence of that court.
Kytmanovsky District, Altai Krai, January 2020 — a second prosecution before the same rural district court whose May 2020 judgment was added this morning. 114,000 roubles from two women. Establishing that these were two different offenders, not one case counted twice, took comparing case numbers, UIDs, presiding judges, offence dates, victims, sums and prior-conviction histories. They are different women; both had earlier convictions under the same fraud article.
One judgment from this batch was read and rejected. A woman in Berdsk was convicted of extortion after seizing another by the hair and demanding 2,000 roubles back — money she had handed over days earlier to someone claiming to be a psychic who told her she was cursed. She was the victim of the curse-removal fraud, and the person convicted. The archive records harm done through spiritual authority to those subject to it, and this judgment runs the other way, so it is not a record here.
The Altai record published this morning has also been corrected. Its scope note said the defendant's prior convictions were not detailed; the judgment in fact records a 2011 ten-year sentence and a March 2020 fraud conviction that the Altai Regional Court modified on appeal to a suspended term. That appeal concerns a different case and does not change this one's appellate status, but a reader assessing the defendant should have it.

Effect on published figures

Russia rises to forty-two records and Crimea enters the archive as a separately-attributed jurisdiction. Forty-five judgments on this practice have now been read across Russia and Ukraine, of which forty-four qualified.

July 29, 2026 records

Three more Russian judgments: 12,004 roubles, 221,000 roubles, and a shop bolted from the inside

Three judgments from Altai, Vladimir and Nizhny Novgorod oblasts, spanning the full range of what this fraud takes and how far it will go.

Kytmanovsky District, Altai Krai, May 2020. The same opening line as the large cases — a curse is on you, and I can cure it — delivered on a farmstead to two women. The itemised list of what was taken is the record: a dental bridge of five gold crowns, a child's gold cross, three one-year-old laying hens, a litre of jam, two jars of pickles, fifteen eggs. Total 12,004 roubles, which the court found significant given the victim's circumstances. Two years, suspended; recidivism was an aggravating circumstance.
That case sits at the opposite end of the same practice from the Tver judgment added today, where one victim lost 8,182,000 roubles. The method does not change with the size of the target. What changes is what the target has.
Kovrov, Vladimir Oblast, May 2020. A woman approached outside a shop and told a curse was on her, and that removing it required her money to be 'cleansed'. She collected 26,000 roubles from home, withdrew 195,000 from her Sberbank account, and followed the defendant to an area among garages, where she was handed back advertising leaflets prepared in advance to look like banknotes.
One detail from that judgment is recorded precisely because it cuts against the archive's own framing: the prosecutor withdrew from the charge the allegation that the victim was defenceless and helpless by reason of age, on the ground that the file held no evidence for it. The court still treated her advanced age as raising the offence's seriousness. Both things are true and the record says both.
Vetluzhsky District, Nizhny Novgorod Oblast, offence October 2019. The smallest loss of the three and the most coercive. Two women told a shop assistant, alone behind her counter, that a curse was on her. One asked her for paper and matches and performed a rite with them; the other bolted the shop from the inside with a metal latch and drew the curtains so no one could enter or see in. They then told her she faced death and that they were ready to help. Her gold ring, worth 2,400 roubles, was taken on the basis that it had to be buried in a cemetery at night.
That judgment is only partly anonymised — one defendant is masked, the other named in the operative part. The archive names neither. Where a court's own redaction is inconsistent, the archive applies the stricter rule rather than the court's.

Effect on published figures

Russia rises to thirty-nine records, and Altai Krai enters the archive as a new region. Forty judgments on this practice have now been read across Russia and Ukraine.

July 29, 2026 records

Nine counts, six victims, and a curse that was manufactured on the spot

The largest judgment the archive holds for this practice, and a correction to a superlative published yesterday.

Tver, July 2020: three women convicted on nine counts committed against six people over four and a half years. The recorded losses total more than 16.7 million roubles — one victim alone lost 8,182,000, and a further 1,301,000 was taken from him by two of the defendants acting together.
That supersedes a figure published here yesterday. The Stavropol entry described 1,364,000 roubles as the largest single loss the archive had recorded; this judgment is six times larger. The earlier statement was accurate when written and is now wrong, which is what happens when a corpus grows — it is corrected here rather than quietly left standing.
What makes this judgment worth reading in full is that it shows the belief being manufactured, not merely exploited. The victim brings an egg; the defendant breaks it into a bag behind her back and shows the contents, and the victim sees something dark 'resembling fur or a feather'. For another victim, a page torn from a notebook he had brought with him developed the words 'A curse has been placed on you', together with a cross.
The escalation is recorded in the court's own sequence: all the money the victim had, then all her gold jewellery, and then — the judgment says the defendant acted 'realising that the victim had no money or gold items remaining' — requests that she buy food and other goods. The demand did not stop when the victim was emptied. It changed form.
Sentences were three years six months and three years three months in a general-regime colony, and three years one month suspended. Civil damages were awarded to all six victims, including 6,682,000 roubles against one defendant and 1,299,000 jointly and severally against two.
The judgment is anonymised, as Russian judgments routinely are. The record identifies it by case number and court and makes no attempt to name anyone.

Effect on published figures

Russia rises to thirty-six records. Thirty-seven judgments on this practice have now been read across Russia and Ukraine, and the archive's stated maximum single loss is corrected from 1,364,000 to 8,182,000 roubles.

July 29, 2026 records

1,364,000 roubles from one woman, and a Polish register that turned out to be empty

The largest single loss the archive has recorded for this practice, and an honest negative result from a new country's court database.

Stavropol, August 2019: the court found the defendant and an unidentified accomplice 'decided, under the pretext of removing a curse, to steal the property of' a woman, 'distributing the criminal roles between themselves' — the other person's assigned job being simply to start the conversation. From that one victim they took 1,364,000 roubles.
That is the largest single loss in the thirty-six judgments the archive has now read on this practice, and it was charged on the especially-large-scale limb of the fraud provision. Convicted on four counts including one attempt; three years in a penal colony.
Poland was tried as a new source and produced nothing. Its national judgment database is open and large — 539,099 decisions — but the pattern is not findable there. This is recorded rather than passed over, because a negative result from a working database is information.
Two things about that search are worth stating, because they would mislead anyone repeating it. The site's own search form submits without carrying the query and returns every judgment in the system, which looks like a successful search and is not. And the database matches word stems, so a search for fortune-telling returns judgments containing the ordinary verb meaning 'to bode' — the first criminal case inspected turned out to be an attempted homicide in which someone remarked that events 'boded only further problems'.
A count of search results from that register is therefore not a count of cases. The archive's method depends on searching multi-word phrases describing conduct, and Poland's search does not support them: a two-word query returns zero.

Effect on published figures

Russia rises to thirty-five records. The Polish register is recorded as tested and empty for this pattern, with the specific traps noted so it is not re-tried on the same assumption.

July 29, 2026 records

Three convictions for the same fraud, three suspended sentences

Two more Russian judgments, containing the largest single sum the archive holds for this practice and a defendant convicted of it for the third time.

Nizhnekamsk, Tatarstan, June 2018: the court recorded that the defendant had 'twice previously been convicted of the theft of the property of others by deception', that 'after conviction she drew no proper conclusions for herself, did not take the path of correction', and that 'while serving suspended terms of punishment she again committed, by an analogous method' the same offence.
The plan is the part worth quoting. She and another person 'developed a plan for committing the offence, settling as the method of deception the presence of a curse and of deadly illnesses'. The diagnosis and its cure were not improvised on meeting the victim — they were the agreed instrument, written into the scheme.
The loss was 494,500 roubles. She received four years, suspended, with a four-year probation period. A third conviction for the same conduct, a third sentence served outside prison. The record states that sequence because the judgment does, and does not editorialise about the court's reasoning, which it does not summarise.
Samara, April 2019: two women 'planned their actions in advance, distributed roles within the criminal group and settled the method of committing the offence — the theft of citizens' money by fraudulent means under the guise of providing healing services'. One approached an elderly woman in the street and 'introduced herself as Elena Ivanovna, a famous psychic'.
They took 720,000 roubles, the largest single figure the archive holds for this practice, and a further 200,000. Both sentences were suspended. Neither went to prison.
Reading these alongside the Nizhny Novgorod case added earlier — fourteen counts, five and a half years — the range is stark. The same conduct has drawn an 850-rouble fine, four years suspended after two previous convictions, and five and a half years in a penal colony. The archive records what each court decided without reconciling them.

Effect on published figures

Russia rises to thirty-four records. Thirty-five judgments on this practice have now been read across Russia and Ukraine.

July 29, 2026 records

Fourteen counts, fifteen elderly victims, five and a half years

The largest case of this kind the archive holds: a woman convicted in Nizhny Novgorod of a sustained campaign against elderly people living alone.

The court found she 'committed thefts of the property of others by deceiving elderly citizens' across a period, and — the phrase that distinguishes this case — that she acted 'with a view to ensuring conditions for entering the dwellings of lonely elderly people for the subsequent theft of property by deception'.
That is not a description of someone taking chances as they arose. It is a finding that getting inside the homes of isolated old people was the object she worked towards, and the curse was the way in.
Fourteen counts, fifteen numbered victims, individual losses including 91,010 and 120,000 roubles. Sentenced to five years six months in a penal colony — the heaviest penalty in the thirty-three Russian and Ukrainian judgments the archive has now read.
The court also expressly refused to defer her sentence under a provision that would have allowed it, on grounds the published judgment redacts. The refusal is recorded because the court states it; the circumstance is not described here because the text does not disclose it.
The judgment runs to roughly 200,000 characters. The record summarises its opening findings and its operative part, and states plainly that it does so — no total loss is claimed across all fourteen counts, because the judgment does not total them in a form that can be quoted with confidence.

Effect on published figures

Russia rises to thirty-two records. Across Russia and Ukraine the archive now holds thirty-three judgments on this practice, with sentences ranging from an 850-rouble fine to five and a half years' imprisonment.

July 29, 2026 records

Children sent home for their families' gold, and a court that stopped to define what a curse is

Two Russian convictions. In one the people deceived were children, who were sent to fetch money belonging to adults. In the other, targeting the elderly was written into the plan before anyone was approached.

Yaroslavl, December 2019: the court found the defendant approached two children she did not know in the street and told them 'information not corresponding to reality, that a curse lay on their families'. She led them to a sparsely populated spot behind two buildings and 'performed manipulations supposedly confirming' it, then told them that to lift the curse they must hand over money and gold belonging to their families. Sums of 25,500, 5,100 and 250 roubles are itemised, with property totalling 35,550.
The deception was practised on children; the loss fell on their parents. Neither judgment nor record makes any finding that the children were harmed.
That judgment does something unusual: it stops to define its terms. It glosses a curse as 'a superstition, a myth about the possibility of a negative magical influence on the part of ill-wishers, and about the possibility of neutralising such influences'. Courts rarely explain the belief they are adjudicating, and the gloss is recorded as the court's, not this archive's.
Vladimir, November 2019: the defendant formed an intention 'to commit the theft of money from elderly citizens by deception, under the guise of removing a curse' and recruited an acquaintance. The two 'divided the criminal roles between themselves in advance', hers being to find elderly people, tell them a curse had been laid on them and offer to remove it. 400 hours of compulsory work; 21,000 roubles lost.
The second woman is not recorded as convicted. Her prosecution was discontinued on reconciliation with the victim, and the archive makes no finding against her — she appears only because the court described the roles they agreed.
One judgment in this batch names its defendant in full where the anonymisation was not applied. That name is not in this archive. Where a court has anonymised a defendant, an imperfect redaction does not license identifying her.

Effect on published figures

Russia rises to thirty-one records. Of the last four judgments read, all four described the practice as planned in advance around a chosen category of victim: the elderly, or children.

July 29, 2026 records

The curse is on your child: two more Russian convictions

Two judgments in which the claim was not that the victim was cursed but that her child was. Both mothers paid; neither defendant went to prison.

Naberezhnye Chelny, Tatarstan, January 2026: a woman was approached outside a bakery while out with her son. The court found the defendant acted 'under the pretext of removing a curse, reliably knowing that she did not possess such abilities', told the mother a curse had been laid on the boy, and offered to help lift it. The mother agreed and met her again the next day, bringing her own mother — who the judgment expressly records knew nothing of the defendant's intentions. The loss was 176,437 roubles.
Tver, December 2019: the same shape. The defendant told a woman that a curse had been laid on her and on her daughter, and took 155,000 roubles across a series of handovers made in a parked car. Two years, suspended.
Across the Russian and Ukrainian records now held, the pretext has been a curse on the victim, on her money, on her family, on her mother, on her son and on her daughter — and in one January 2026 case from Krasnodar, magical rituals to save the life of a relative. The claim adapts to whatever the person in front of them cares about most.
Both sentences here were suspended, and both records say so plainly. Of the twenty-five Russian records now held, most ended in a fine or a suspended term rather than imprisonment.
In both cases the child is recorded because the court found the curse was said to lie on them — that is what gave the claim its force. Neither judgment found any harm to the child, and neither record asserts any.

Effect on published figures

Russia rises to twenty-seven records. A broader search phrase returns substantially more judgments than the one used until now, so this channel has more depth than previously estimated.

July 29, 2026 records

Half a million roubles, and a scheme that named the elderly as its target in advance

Two Russian convictions, including the largest sums the archive has recorded for this practice and a judgment from this January that sets out the targeting in the plan itself.

Krasnodar, January 2026: the court found the defendant and another person 'entered into a prior criminal conspiracy and developed a criminal plan aimed at the theft of money and jewellery by deception from an indefinite circle of persons, including persons of advanced (elderly) age, under the pretext of performing magical rituals to save the life and health of the victim's relatives'.
That sentence is worth reading twice. The elderly were not who the scheme happened to reach — they were written into the plan. And the pretext was not a curse on the victim but the survival of someone she loved.
One woman lost 165,000 roubles and another 500,000, the latter recorded in the judgment as severe given her very small means. Three years for the two offences, aggregated with an earlier sentence to three years six months in a penal colony.
Krasnoyarsk, May 2020: a conversation outside a market pavilion. The defendant told a stranger a curse had been laid on her and a rite was needed to cure her, offering healing services she had no ability to provide. The payments went 200 roubles, then 2,000, then 44,000, then 20,000, 94,000 and 150,000 — 310,200 in total.
The individual figures are listed rather than only the total, because the escalation is the thing being documented. A first payment of 200 roubles is not a sum anyone would refuse to risk. That is how it starts.
Neither defendant is named, as Russian judgments are anonymised. One of these judgments does contain a surname in its operative part where the anonymisation was evidently not applied; this archive does not reproduce it and has not used it to identify anyone.

Effect on published figures

Russia rises to twenty-five records. Across Russia and Ukraine the archive now holds twenty-three judgments describing this single practice, read in the courts' own words.

July 29, 2026 records

The second half of a case already here, and a woman working a hospital's grounds

One new record, and one existing record completed rather than duplicated — the second prosecution arising from an episode the archive already held.

The Boiany case in Chernivtsi Oblast, added earlier today, described one woman convicted of planting notes outside a shopkeeper's kiosk so that a curse would seem to explain her troubles. That judgment referred throughout to a second participant whose case had been severed, and the record said plainly that no finding was being made about her.
That severed prosecution has now been found and read. She was convicted in April 2014 and released on probation. Her judgment describes the same events from the other side and divides the roles identically — one scattered the notes, the other received the thanks in the victim's hearing.
It has been added to the existing record as a second proceeding rather than published as a new record. One episode, two convictions, one entry. Adding it separately would have shown two cases on the map where only one thing happened.
One limitation is stated in the record: the two judgments use different placeholder codes for the participants, so a reader cannot match a code in one to a code in the other. The record therefore describes what each role involved rather than claiming to track individuals across the texts.
Rozhyshche, Volyn Oblast, December 2010: a woman was convicted of working the grounds of a district hospital — telling people there that they and their relatives were in danger and promising to heal them by divination. She took 2,000 hryvnia from one and 60 from another. Sentenced to three years, with two months of an earlier unserved sentence added, making three years and two months in custody.
That is the second-heaviest sentence in the Ukrainian set, and the second case here where the approach was made where people were already worried about someone's health. The record notes the hospital and its staff had no involvement.

Effect on published figures

Ukraine rises to fifteen records covering sixteen convictions. Eighteen judgments have now been read from the register.

July 29, 2026 records

When the victim makes the first move: two cases where reputation did the work

Two more Ukrainian convictions, both differing from the street approaches recorded so far. In these, the women were not stopped by a stranger — they went looking for help, and one of them brought her child.

Khmilnyk, Vinnytsia Oblast, August 2016: the court found the defendant decided to defraud residents of the town and, to that end, arranged through someone else to rent a house there, arriving days later. A woman then telephoned her to say she was having trouble in her family and with her small son's health.
The defendant told her 'that she possessed superhuman abilities' and that a curse-removal rite was needed, and invited her to come the next morning and bring the boy. When they arrived she said the child was ill because a curse lay on the family, and that she could lift it. She took 50 hryvnia and food, then continued the rite using the child's own belongings, taking further sums of 625 and 950. The record notes the boy was the subject of the false claim of illness and was present, and that no physical harm to him was found.
The renting of the house is what marks this case. It was taken before any victim existed, so the practice here is not opportunism but a small business with premises.
Nadvirna, Ivano-Frankivsk Oblast, December 2013: the court found the victim sought the defendant out 'in order to learn about her future, since it had become known to her' that this woman told futures and could remove curses. They met in the street and walked into a hospital, and in the corridor the defendant told her a curse had been laid on her family and her mother. Fined 1,700 hryvnia.
One figure in that judgment needs care and the record states it plainly: 22,940 hryvnia appears in the text, but it is the bail the defendant had posted — not a loss, not a fine, and not compensation. It would be easy to report it as the sum taken.

Effect on published figures

Ukraine rises to fourteen records. Sixteen judgments have now been read from the register, covering courts in thirteen oblasts.

July 29, 2026 records

Planned by telephone: two women who arranged a curse-removal scheme before choosing a victim

Two more Ukrainian convictions. One shows the practice organised in advance rather than improvised, and both defendants went to prison — which most of these cases do not.

Tarashcha, Kyiv Oblast, July 2011: the court found two women agreed a scheme by telephone on one evening, met at one of their homes the next morning where 'they jointly drew up a plan of their criminal activity', then took a taxi to a village in another district and entered a married couple's house 'under the pretext of a rite of divination and the removal of a curse'.
The sequence matters. The plan came first and the victims were chosen afterwards. This was not someone spotting an opportunity in the street but a scheme organised around the pretext, with travel arranged to reach a household neither of them knew.
Both were convicted of fraud by prior conspiracy and both received prison terms — eighteen months and two and a half years — with damages awarded against them jointly. Of the Ukrainian records held here, most ended in a fine or a suspended sentence; this one did not.
Staryi Pochaiv, Ternopil Oblast, May 2017: a woman called at a village house, introduced herself as someone who tells the future and can remove curses, told the householder a curse had been laid on her, and telephoned the next day to continue the deception. Fined 510 hryvnia.
That conviction came from the same court as one already in the archive, six weeks apart, in almost identical language. The distinction was checked carefully before adding — different case number, different victim, different date, and a house call rather than a café. Two judgments reading alike is a feature of settled legal phrasing for a familiar offence, not evidence of duplication.

Effect on published figures

Ukraine rises to twelve records. Fourteen judgments have now been read from the register, from courts in eleven different oblasts.

July 29, 2026 records

Planting the curse: a case where the belief was manufactured, not merely exploited

Two more Ukrainian convictions. One of them documents something none of the others do — the fraud began by creating the evidence that a curse existed at all.

Boiany, Chernivtsi Oblast, March 2013: two people drove to a village, learned that a shopkeeper was having domestic difficulties, and prepared for her a reason. The court found the defendant planted 'notes and other objects in the manner of divination' by the entrance to the woman's trading kiosk, which 'were to serve as a basis for various suspicions arising' in her — suspicions that her business and her family's wellbeing were being blocked.
The next day the two walked into the kiosk together, and in the shopkeeper's hearing the defendant thanked her companion 'for supposedly having helped her in the removal of a curse'. The court found this created a belief in that person's powers 'which contributed to' the shopkeeper approaching her for help herself. She was never solicited. She asked.
That completes a picture the earlier records only sketched. Other judgments describe a second person brought along to vouch for the healer; this one shows the physical evidence being planted first, so that when the victim hears the endorsement she already has her own reason to believe. Conviction for fraud by prior conspiracy; one year's restriction of liberty, already covered by time in custody.
Lviv, April 2014: a woman approached shoppers in a shopping centre claiming she 'possessed the skills of removing a curse', took a gold chain and cross worth 1,500 hryvnia, and — the court found, knowing she would not return them — convinced the owner they would come back after the rite. Two years' imprisonment was imposed but then suspended with three years' probation, which the record states plainly so the sentence is not read as time served.
Two further documents were read and not used: both were procedural rulings, one authorising access to bank records during an investigation and one on temporary access to documents. Neither is a conviction, and the archive does not record investigations.

Effect on published figures

Ukraine rises to ten records. Filtering the register to verdicts only leaves about forty judgments on this exact offence phrasing, of which roughly a dozen have now been read.

July 29, 2026 records

A four-year sentence, a male victim, and a case nobody was ever caught for

Three more Ukrainian judgments read, two added. The third was closed without anyone being identified, which is worth reporting because it is what most of these cases probably look like.

Kamianske, Dnipropetrovsk Oblast, January 2021: the heaviest sentence in the Ukrainian set. The court recorded that the defendant was already convicted of offences against property and had 'drawn no proper conclusions for herself'. She approached elderly women in the street, told each 'that a curse lay upon her, for the removal of which it was necessary to perform a corresponding ritual', went with them into their own flats, and took their money while they were — in the court's words — bewildered and did not understand a crime was being committed against them. Four years nine months.
That figure needs care and the record states it plainly: it is a final aggregate covering a robbery count, a fraud count and an earlier unserved sentence. It is not the penalty for the curse-removal fraud alone.
Chernihiv Oblast, June 2018: a variation worth noting. The victim was a man, and the claim was not that a curse lay on him but that it lay on the money in his house — so he was told to bring out everything he had, for manipulations and prayers to be said over it. He brought 1,000 hryvnia and 20 euros. The record notes the victim's sex only so readers do not infer this is practised solely against women.
Poltava Oblast: a case from 2015 in which a woman reported losing gold jewellery, 8,400 hryvnia, 20,000 roubles and a phone to someone offering divination and curse removal. It is not added, because in 2021 the court closed the proceedings — nobody had ever been identified, and the limitation period had run out.
That closure is included here rather than passed over silently. This archive only records adjudicated outcomes, so cases where no one is caught never enter it. The convictions on this site are the fraction of this practice that ended in court, and the Poltava file is a reminder of the far larger number that do not.

Effect on published figures

Ukraine rises from six records to eight. Of eleven judgments read from the register so far, seven produced records, one was a duplicate and three failed on the evidence.

July 29, 2026 records

The rite as sleight of hand: real money swapped for souvenir notes

Two more Ukrainian convictions. One shows the practice at its smallest — a doorstep in a village — and the other at its most elaborate, where the ritual existed to create the moment when the money could be switched.

Kyiv Oblast, January 2023: an aunt and her niece approached a woman near her building and told her they could resolve her health problems by removing a curse. She went home, collected 1,620 US dollars and 8,200 hryvnia — about 67,000 hryvnia in total — and brought it back to them.
What happened next is why the rite mattered. The court found the aunt began performing it, speaking aloud as she did, and while the woman's attention was on the ritual she substituted souvenir banknotes for the real money, put them in a bag, placed it in the woman's handbag and told her to go home and tell nobody about the rite. The niece stood nearby and kept watch. The instruction to stay silent is part of the mechanism, not an afterthought.
Both women were convicted and each fined 51,000 hryvnia. The record notes the money was returned, so it is not read as an unrecovered loss, and that both entered plea agreements — so the findings are agreed rather than contested.
Zhytomyr Oblast, January 2010: a woman travelling a village by hired car, ostensibly to buy pumpkin seed, nuts and beans from residents, offered at one house to tell fortunes and remove a curse. She was invited inside and took 2,000 hryvnia. Fined 1,700.
That case is worth keeping beside the other. The same claim — I can lift the curse on you — worked as an opening line on a doorstep in 2010 and as cover for a sleight-of-hand switch in 2023, in sums two orders of magnitude apart.

Effect on published figures

Ukraine rises from four records to six. A search on the exact phrase Ukrainian courts use for this offence returns 97 judgments; eight have been read so far.

July 29, 2026 records

Two more Ukrainian judgments, and courts describing the same offence in near-identical words

Working through Ukraine's court register produced two further convictions and one duplicate. The duplicate is worth mentioning, and so is a detail visible only when several judgments are read together.

Poltava, July 2010: a woman approached a stranger in Pavlenkivskyi Park and took her cash and gold 'under the pretext of divination and the removal of a curse', having convinced her the items would be returned after the rite. A gold chain, two rings, a bracelet, earrings, a brooch and 274 hryvnia — 5,924 in total. She was sentenced to eighteen months' imprisonment, though that term covers a second, unrelated theft count as well, which the record states rather than letting the sentence stand for the fraud alone.
Kremenets, Ternopil Oblast, March 2017: a woman entered a café-bar and introduced herself as 'a person who tells about the future and can remove a curse by divination', then told two women in turn that a curse had been laid on them. Convicted on two counts; the sums were small — 370 and 1,967 hryvnia — and the sentence was a fine of 1,700, following a reconciliation agreement with both victims.
A third result was checked and not added: it was an earlier procedural stage of a case already held here, same case number and same facts. Reading it was not wasted — confirming a duplicate before writing is how the archive avoids counting one prosecution twice.
The detail that only appears across several judgments is the wording. The Kremenets and Bila Tserkva courts, in different regions seven years apart, describe the offence in nearly identical phrases: the defendant introduced herself as someone who could 'remove a curse' by divination, then acted 'taking advantage of the trusting features of her character'. That is not two judges coincidentally reaching for the same words. It is a recognised offence pattern with settled legal language, which tells you the courts see it often.
None of these defendants is named. Ukrainian judgments are anonymised, so each is identified by case number and court.

Effect on published figures

Ukraine rises from two records to four. The curse-removal fraud pattern is now documented from the courts' own judgments in six countries.

July 29, 2026 records

The same routine, in Ukraine — and a fine of 850 against 15,097 taken

Ukraine's court register has been opened as a source. The first judgment read from it describes a routine already documented five times over in Russian courts, which says something the individual cases do not.

Bila Tserkva, Kyiv Oblast, July 2021: a woman approached an elderly stranger in the street and introduced herself as someone who could 'remove a curse' by divination and ritual. The court found she acted 'taking advantage of the trusting features of her character' and 'reliably understanding that the latter was a person of advanced age'. She told the victim a curse had been laid on her, then that removing it would take several meetings of manipulations and prayer, each to be paid for.
The payments escalated across those meetings: 500 hryvnia, then 1,000, then 4,000, then 5,447, and on — 15,097 in total. The sentence was a fine of 850 hryvnia. The record states that disparity plainly because the judgment does, while drawing no conclusion about Ukrainian sentencing generally from a single case.
What matters most is that this is the fifth country where the archive has read this exact routine out of a court's own judgment. Approach an older woman in public, name a curse, structure removal as a series of paid sessions. The Russian judgments added a second person who vouches for the healer; this one did not need it.
As with the Russian records, the defendant is not named — Ukrainian courts anonymise published judgments — so she is identified by case number and court instead. The register also publishes the date a judgment took legal effect, which is more than most sources give: this one became final on 19 August 2021.
Two smaller changes shipped alongside. The site's Support tab is now 'Support the Project', which should stop it reading as technical support. And that page carries a progress bar toward the next records milestone, computed from the corpus rather than typed in — currently 1,330 of 1,500.
The progress bar's caption says what a milestone is not: a measure of work done, not of harm in the world. A growth counter on a donation page invites the reading that more records means more abuse, and that inference is wrong — the archive grows where records are published.

Effect on published figures

Ukraine goes from one record to two, and its national court register joins Russia's as a working source. Fifty-six further judgments match the same search.

July 29, 2026 records

Children chained in Chad, and three West African findings

Three more findings from the UN Committee on the Rights of the Child, from countries where the archive held almost nothing. One of them is the most physically explicit finding in the treaty-body material so far.

Chad, 2009: the Committee reported 'that children in Koranic schools are often chained up because they are considered difficult or rebellious and that some children are forced to go out begging on behalf of their Mouhadjir, and are beaten if they do not bring back a certain amount of money'. It made this finding under the heading of torture, and noted that Chad's Penal Code contained no explicit definition of the term. Chad had no records here before this.
Côte d'Ivoire, 2019: children with albinism are 'victims of ritual killings, abductions, abandonment and stigmatization', and — a distinct finding worth stating on its own — 'the violations against child victims are not identified as such for the purpose of data collection'. Where a state does not count a category of victim, the scale of the harm cannot be established. No figure is given here for that reason.
Mali, 2024: the Personal and Family Code sets the marriage age at sixteen for girls and eighteen for boys, and child marriage 'reportedly affects one girl out of two'. The Committee's recommendation is directed at amending that specific law. The same paragraph records that attacks on children with albinism are carried out for ritual purposes — recorded here, but explicitly not attributed to the Code or to any religious authority. Two findings sharing a paragraph do not share a mechanism.
Three countries were examined and turned down. Niger's finding was about customary law generally, with no instrument named. Cameroon mentioned children with albinism only in a list of groups facing discrimination, with no ritual-killing finding — and its Council of Imams appears as a partner the state is urged to work with, not as a perpetrator. Bangladesh, examined earlier, was the same shape.
That is now six rejections against this rule in two days. The rule was adopted because religiously-governed states were under-recorded, and the obvious way to abuse it is to reach for it whenever a country is Muslim-majority. Each refusal is recorded so the pattern of restraint is as visible as the pattern of inclusion.

Effect on published figures

Institutional records rise to 159. Chad and Côte d'Ivoire go from no records to one each; Mali from one to two. Ten records now rest on the fused-authority rule.

July 29, 2026 records

Two of the archive's thinnest records now rest on UN findings instead of a press report

Rather than add new records for practices the archive already covered, two existing entries were upgraded. Both were among the weakest here: a single source each, no proceedings, no statement of scope.

Ghana's trokosi record — girls placed with a shrine as atonement for an offence committed by a relative — rested on one article from a women's development fund. It now also carries the UN Committee on the Rights of the Child's 2015 finding, which establishes something the original could not: trokosi has been a criminal offence in Ghana since 1998, and as of May 2015 'not one case has been reported and investigated'. Seventeen years of a crime with no prosecutions.
Benin's voodoo convents record rested on a US trafficking report. The Committee's 2016 finding adds three facts it did not establish: children sequestered in these convents are 'deprived of education and contacts outside of the convents', and are 'subjected to ill treatment, including scarification rites and sexual abuse'.
Both records now say what they are not. Vodun is a living religion with millions of adherents, and the overwhelming majority of its practitioners are not implicated by a finding about children held in particular institutions. Neither record names an individual or a specific convent or shrine, and nobody has been prosecuted in either country.
The Ghana record also now states plainly that the figure in its title comes from the original reporting and is not a UN finding, so a reader does not attribute a number to a body that never gave one.
A note on why this was done as an upgrade rather than as two new records: the archive already covered both practices, and adding parallel entries would have inflated the count while making the same subject harder to find. Where a better source turns up for something already held, the record should get better rather than the total getting bigger.

Effect on published figures

No change to the number of records. Two of the corpus's weakest-sourced entries now rest on treaty-body findings, with proceedings and scope statements they previously lacked.

July 29, 2026 records

Children accused of witchcraft in the Congo, and graves opened in Malawi

Two more findings from the UN Committee on the Rights of the Child, both about children harmed for what they were believed to be rather than by anyone with authority over them.

Democratic Republic of the Congo, 2017: the Committee expressed grave concern at reports of 'an increasing number of murders of, and incidents of ill-treatment and violence against, children accused of witchcraft and the impunity faced by perpetrators linked to certain churches'. It urged the state to criminalise the persecution of children accused of witchcraft and to direct prevention work at parents and religious leaders.
The phrase 'linked to certain churches' is the Committee's, and it does not say which. Neither does this archive. The record is explicit that it is not a finding against any denomination or congregation, and that the great majority of Congolese churches are not implicated. The Committee also names poverty as a root cause and directs prevention at parents, which the record keeps so the practice is not attributed to churches alone.
Malawi, 2017: the Committee recorded 'the significant number of cases of abduction, ritual killings and exhumation of remains of children with albinism' — bodies taken from graves after burial — along with 'reportedly lenient sentences handed down to perpetrators' and inadequate enforcement of the laws meant to protect these children.
It also found high school drop-out rates among children with albinism, caused by fear of attack and by being prohibited from wearing sun protection clothing. A child kept out of school by fear of being killed for her body, and unable to protect her skin when she does attend, is a second harm running alongside the first.
The archive already holds a 2022 Malawian conviction in which three men received effective 155-year terms for a ritual murder. That is a different thing from this: one is an individual prosecution, the other a national finding five years earlier. The record says so, and notes that the Committee's remark about lenient sentences describes what was reported to it in 2017, not that later case.
Bangladesh was examined and not added. The Committee found child marriage highly prevalent there, but identified no religious provision or authority as the mechanism — it recommended awareness work aimed at religious leaders among others, which is not the same as finding they caused it. Prevalence without a named mechanism is not a record here.

Effect on published figures

Institutional records rise to 156. Both new records are inverse ones — children harmed because of what belief attributed to them — and they are counted separately from records of authority abuse.

July 29, 2026 records

Children on death row, and children punished by being left in the sun

Two findings from the UN Committee on the Rights of the Child. One concerns a penalty that can reach children because religious law defines childhood differently. The other concerns children hunted for their body parts — and what happened to them in the shelters built to protect them.

Nigeria, 2010: the Committee reported around forty prisoners on death row for crimes committed before they were eighteen, and warned that the mandatory death penalty under sharia penal codes, including Hadd punishments, 'could be imposed on children under sharia jurisdiction' — because those codes define adulthood by puberty rather than by age, and most northern states had not adopted the Child Rights Act that would forbid it.
The record is careful about what it does not say. It does not assert that any child has been executed. The Committee's finding is that the penalty could reach a child given how childhood is defined there. The forty prisoners are a figure reported for Nigeria as a whole, not for sharia courts, and the record says so.
This is also the second time the archive has examined Nigeria under the new fused-authority rule. A separate finding about a constitutional provision on nationality was rejected, because no religious warrant was identified in it. This one names the sharia penal codes directly. The difference between the two is exactly the line the rule draws.
Tanzania, 2015: the Committee was 'extremely alarmed' at the killing of children with albinism for ritual purposes, and found that prosecution is hampered 'by fear and the reported complicity of some State authorities'. It directed the remedy it recommended at the practitioners it called 'witch doctors'. This is an inverse record: the children were not under anyone's spiritual authority — they were killed because a belief about albinism made their bodies valuable.
The second half of that record is harder to read and belongs with the first. The Committee found the shelters built to protect these children were not temporary as intended, were unhygienic and overcrowded, separated children from their families, and that inside them children 'are punished, especially by prolonged exposure to the sun, leading to the rapid development of skin cancer'. Children with albinism, punished with sunlight. They were also subject to threats and sexual abuse. The state's answer to the killings became its own harm.
The archive already held albinism killings in South Africa, Malawi and Mozambique but not Tanzania, where the practice is most documented. The term 'witch doctors' is kept in quotation marks throughout because it is the Committee's word, not the archive's, and no claim is made about African traditional healing generally.

Effect on published figures

Institutional records rise to 154. Nine records now rest on the fused-authority rule. Tanzania joins the albinism records, which now span four countries.

July 29, 2026 Correction

A record was removed a day after it was added, and the rule tightened so it cannot return

Yesterday's new rule about religiously-governed states has already been stretched past what it can bear. A Yemen record built on that stretch has been withdrawn, and the rule now has a fifth limit.

The rule says a record may only rely on it if it names the constitutional or legal provision supplying the religious authority. The Yemen record named an absence: Yemen sets no minimum age of marriage, so the argument ran, the question defaults to religious and customary law, and the warrant is 'supplied by absence rather than by enactment'.
That is rejected for two reasons. It inverts the requirement — a record must cite which provision does the work, and an absence has no provision to cite. And it has no stopping point: on that reasoning every gap in every country's legislation becomes religious authority, which would make the rule impossible to argue against and hand critics precisely the objection its safeguards exist to prevent.
Checking the underlying document settled it. The UN committee's report on Yemen contains no reference to a personal status code, to sharia, or to any reservation. There was no instrument to cite because the committee never identified one. What it did identify was a stalled parliamentary process, marriages arranged by families, and girls forcibly married to an armed group — grave harm, but not the mechanism this rule was written for.
The Syria record added at the same time stands, and the contrast is the point. There the committee names the Personal Status Code of 1957 and a specific judicial discretion it confers to lower a girl's marriage age to thirteen. A citable instrument is the line.
The rule now carries a fifth limit stating this directly: a legislative vacuum is not a religious warrant, however grave the resulting harm. Where a state has failed to legislate, the finding to record is that failure, in whatever category the actual mechanism supports. An automatic check now rejects any record arguing otherwise.
Also added: Mauritania, where the committee found boys in Qur'anic schools 'forced to beg on the street for the economic gain of their teachers' and called for talibé children to be removed from those teachers' control. That record does not use the new rule at all — a Qur'anic teacher is a religious figure in his own right, so the ordinary test is met without it.

Effect on published figures

One record removed, one added. The archive's newest rule has been narrowed a day after adoption, which is the correct direction of travel for a rule that widens what can be recorded.

July 29, 2026 Correction

Audit of every published statistic, after two of them turned out to be wrong

Having corrected the country counts twice in one day, the sensible next step was to check whether the other published figures had the same problem. Most did not. Two more faults were found, and one of them was in the fix itself.

Seven aggregate statistics are published in the archive's data files: by tradition, by type of case, by decade, by country, by record class, by evidence status and by pretext category. Six of them are backed by fixed lists — a record's tradition must be one of the declared traditions, and so on — and all six check out. Every category used exists, every declared category is used, every one is labelled, and each sums exactly to the number of records.
The one that broke was the one derived from free text. That is the lesson worth keeping: the country figure was wrong twice today, and it was the only statistic not backed by a fixed list of allowed values.
The first new fault was in this morning's own fix. A record covering proceedings in both the United States and India was left with a country of 'United States and India' — a bucket that was itself nonsense, complete with the double space where a bracket had been removed. It is now attributed to neither country. Being counted in the total but assigned to no country is the honest answer for a case that genuinely spans two jurisdictions; the country figures accordingly sum to one less than the record total, and the API documentation now says so.
The second was worse, because it meant the archive was publishing two different answers to the same question. The analytics data had its own separate copy of the country logic, which stripped a bracketed note but not one containing a comma — so that file still listed 'Michigan', 'Haryana' and 'Tyne and Wear' as countries, and called the United States by a different name than the statistics file did. There were four copies of this logic in total. They now all use one.
Two further checks were added, bringing the suite to 49: no country bucket may name two jurisdictions, and any record left without a country must genuinely be multi-jurisdiction rather than dropped by accident. The second has a ceiling, so if the number of unattributed records starts climbing it fails rather than quietly growing.

Effect on published figures

No records were added, removed or altered by any of this. The tradition, type, decade, class, evidence and pretext figures were correct and remain unchanged. The country figures are now consistent across both places the archive publishes them.

July 28, 2026 Correction

Forty-four records were filed under countries that do not exist

This morning's fix to how the archive counts countries turned out to be half a fix. A second, larger fault was sitting underneath it, and it was producing published country figures that were wrong for the United Kingdom, Ireland, New Zealand and the United States.

Many records record where a case was tried as well as where it happened — 'Sedlescombe, East Sussex, England, United Kingdom (tried at Hove Crown Court)'. The archive worked out the country by taking the last part of that text, which meant the country became 'United Kingdom (tried at Hove Crown Court)'. Add another court and you get another country. There were dozens of these.
Worse were the ones where the qualifier itself contained a comma. 'Sirsa, Haryana, India (trial: Panchkula, Haryana)' produced a country called 'Haryana)'. 'United States (national, 37 states and territories)' produced '37 states and territories)'. Those were published, in the statistics file and the downloadable data, as if they were countries.
Forty-four records were affected. The corrected figures are materially different: the United Kingdom rises from 71 to 88, Ireland to 40, New Zealand from a scatter of variants to 21, the United States to 310. The number of distinct countries falls from 168 to 140 — that is invented buckets collapsing into real ones, not countries vanishing.
The fix strips any bracketed or dash-appended qualifier before reading the country. One case is deliberately left unresolved: a record covering proceedings in both the United States and India is not assigned to either, because it genuinely spans two jurisdictions and picking one would be a false precision.
Two automatic checks now run on every build. One asserts that no record derives a country containing a bracket or a digit. The other asserts that no country appears both with and without a qualifier. Both were verified to fail against the old code before being kept — a test that passes either way protects nothing.
This is the second correction to the same aggregate in one day, and the pattern is worth naming: the first fix addressed spelling variants and stopped there, without checking whether the underlying extraction was sound. Finding one class of fault is not evidence that it was the only one.

Effect on published figures

No records were added, removed or altered — only the counting was wrong. Anyone who has taken a per-country figure from this archive or its data files should take it again.

July 28, 2026 records

The new rule applied to Israel and India, which is the test of whether it is a rule or a target

This morning's rule change was written with a requirement that it apply evenhandedly. Two records now test that: one concerning Jewish religious divorce in Israel, one concerning religious personal law in India — a secular republic.

Israel, 2017: under Israeli law a Jewish marriage can only be dissolved by a get, a religious bill of divorce that the husband must grant of his own will. If he refuses, his wife cannot remarry, indefinitely. The UN Committee on the Elimination of Discrimination against Women listed among forms of gender-based violence 'psychological violence and abuse by husbands against their wives during divorce proceedings through the withholding of consent to divorce (also known as get abuse)'.
The Committee also found that women who cannot obtain a get 'continue to suffer stigma and discrimination when they bear children by men other than their husbands' — the consequence of remaining married to a man who will not release them. The religious warrant is Israel's reservation to the treaty's article on family law, which the state explains by reference to religious communities' autonomy over their own practice.
The record states what Israel has done as well as what it has not: the Committee expressly welcomed a 2012 amendment to the law enforcing rabbinical divorce judgments and the strengthening of the family courts. Sanctions against husbands who refuse do exist.
India, 2014: the same Committee recorded that 'judges often authorize marriages of underage girls based on Muslim personal laws' and that no law requires the registration of marriages. It found the coexistence of separate family-law systems for each religious community 'results in deep and persistent discrimination against women'.
India is a secular republic, and the record says so. The finding is about a plural family-law structure inside a secular constitutional order — which is exactly why the rule is written around where legal authority comes from rather than around a country's official religion. A rule keyed to whether a state calls itself religious would have missed this entirely.
Neither record names anyone. No judge, husband, rabbi or court is accused; both concern structures and documented patterns. And in the India record the Committee's own framing is preserved — it says it is concerned at reports that judges do this, which is a characterisation of evidence before it rather than a judicial finding, and the record says so.

Effect on published figures

Institutional records rise from 150 to 152. Seven records now rest on the fused-authority rule, across Islam in five countries and Judaism in one — which is the evenhandedness the rule requires, demonstrated rather than promised.

July 28, 2026 records

Egypt: the official who performs the marriage, named as the perpetrator

A new source for the archive — the UN committee on women's rights — and with it a record that names not a father or a husband but the religious official who solemnises the marriage.

In Egypt a ma'zoun is the religious official authorised to contract and register a marriage. In 2021 the UN Committee on the Elimination of Discrimination against Women recommended that Egypt 'impose penalties on ma'zoun as perpetrators for contracting such marriages or failing to report them', in cases of child and forced marriage.
The word perpetrators is the Committee's own. It is a treaty body identifying the officeholder who performs the ceremony as responsible for it — not as a bystander who should have reported something, but as the person whose act makes the marriage real. A child marriage is an intention until someone with religious authority solemnises it.
This record does not rely on the fused-authority rule adopted this morning, and does not need to. A ma'zoun holds religious office in his own right, so the archive's ordinary test is met without it. That distinction is stated in the record so the two categories are not blurred.
It also states clearly what it is not. No individual ma'zoun is named or accused; the Committee identifies a category of officeholder, not a person, and the record must not be read as a finding against anyone. Nobody was prosecuted. No figure is given for how many such marriages occurred, because the Committee gives none.
Two cases were examined and turned down. Yemen has no minimum marriage age at all, majority at fifteen and criminal responsibility at seven — but the Committee attributes the delay to disagreement in Parliament, and the marriages it describes are arranged by families or forced by an armed group. That is grave harm with the wrong mechanism: not state religious authority, and this archive does not record family-only cases. Nigeria's constitution contains a provision that the Committee says legitimises child marriage by treating any married woman as an adult — but it frames this as discrimination in nationality law, with no religious warrant cited.
Both refusals matter more than they look. The new rule was adopted precisely because religiously-governed states were being under-recorded, and the temptation is to reach for it whenever a country is Muslim-majority. The rule's first limit says fusion shows religious authority was available, never that it was used. Yemen and Nigeria are what that limit looks like in practice.

Effect on published figures

Institutional records rise from 149 to 150. The archive now draws on two UN treaty bodies rather than one, which matters because they reach different mechanisms — the children's committee reaches age-of-majority law, the women's committee reaches marriage registrars and family law.

July 28, 2026 records

Somalia and Pakistan: when the law itself decides a girl stops being a child

Two more findings from the UN Committee on the Rights of the Child. Both concern the same question — at what age does the law stop treating a girl as a child — and in both the answer is set by religious law rather than by a number.

Somalia, 2020: the constitution defines a child as anyone under eighteen, but the Committee found that 'Sharia law and traditional customs dictate that childhood ends at physical maturity somewhere between 12 and 14 years of age'. Since the constitution makes marriage lawful once a party has reached maturity, that religious definition is what determines when a girl may be married.
The Committee was also deeply concerned about female genital mutilation in Somalia, 'in particular involving girls between 5 and 10 years of age', and found the limited progress in eradicating it was 'notably due to resistance of traditional and religious leaders'. That is religious authority named as the obstacle to protecting small children.
The Somalia record states clearly what it does not claim. Genital cutting is not exclusive to any religion and long predates Islam in the region; the record makes no claim that it is doctrinally required, and notes that the Committee itself recommends working with religious leaders rather than treating them all as obstacles. It also records the anti-cutting bill and Puntland's zero-tolerance policy, so the finding is not read as unrelieved failure.
Pakistan, 2016: the Committee found that the Zina and Hadood Ordinances of 1979 'provide a definition of a girl child that only covers girls up to the age of 16 years or puberty' — so a girl ceases to be a child in law at sixteen or at puberty, whichever comes first. It also recorded concern about sharia law applied under a 2009 regulation across large areas of Khyber Pakhtunkhwa, and about the resulting inconsistency between secular and sharia definitions of who is a child.
Both are records about legislation, not about individuals. Nobody was prosecuted, no court adjudicated anything in them, and no person is named. Neither is a finding about Islam or about Muslims — each concerns specific enacted instruments, assessed by the treaty body those states report to.

Effect on published figures

Institutional records rise from 147 to 149. Somalia goes from zero records to one. Five records now rest on the fused-authority rule adopted this morning, across four countries that between them held one record before today.

July 28, 2026 records

Afghanistan and Iran: three countries that had no records this morning now have one each

Two more findings from UN treaty bodies, admitted under the rule adopted today. Both concern countries where the archive previously held nothing at all — and in one of them, the harm the Committee identifies is that abused children are prosecuted.

Afghanistan, 2011: the Committee on the Rights of the Child found that while authorities systematically failed to prosecute people who sexually abused children, 'child victims are very often considered and treated as offenders, and charged with offences such as debauchery, homosexuality, running away from home or zina'. It found that shame and stigma attach to the child rather than the abuser, and that rape 'has not been clearly defined and separated from the offence of zina' in law — so a child who reports being raped risks being charged.
The Committee also found that girls who are abused face honour killing, the practice of baad, or forced marriage to the man who raped them; that traditional dispute mechanisms perpetuate these practices; and that the impunity of those responsible results in part from the complicity of religious leaders and elders. It recorded that the Penal Code exempts those who commit honour killings from punishment for murder.
Iran, 2016: the Committee found the age of majority set at nine lunar years for girls and fifteen for boys, so that girls above nine lose the Convention's protections, and the legal marriage age set at thirteen for girls. It found this places girls at risk of forced, early and temporary marriage. The religious warrant is named: Iran's reservation to the treaty invokes Islamic law generally, and the Supreme Court relied on that reservation in 2012 to rule that domestic law prevails over the Convention.
One thing was deliberately left out of the Iran record. The same document reports discrimination against Baha'i and Sunni children. That is religious persecution — the inverse of what this archive tracks — and it is recorded separately in a different category, never as authority abuse. The rule adopted today says explicitly that a state fusing religion and government does not turn the persecuted into the abused, and this is the first time that limit has been applied.
Both records state what they are not: findings by a UN committee rather than criminal proceedings, with nobody prosecuted and no individual named. Neither is a finding about Islam or about Muslims — each concerns how one state's institutions operated, as assessed by the treaty body that state reports to. Both describe a specific year, not the countries today. The Afghanistan record notes the Committee also commended Afghanistan for its 2009 law on violence against women, so the finding is not read as unrelieved failure.

Effect on published figures

Institutional records rise from 145 to 147. Afghanistan and Iran each go from zero records to one, joining Saudi Arabia. All three came from a source the archive had never used before today.

July 28, 2026 records

The first record admitted under the new rule: Saudi Arabia, and a religious veto on a marriage-age law

Earlier today the archive changed a rule that had been excluding abuse in states where religious and government authority are the same thing. This is the first record that rule admits, and it comes from Saudi Arabia — a country this archive previously held nothing on at all.

In 2016 the UN Committee on the Rights of the Child examined Saudi Arabia and recorded three findings that together describe the mechanism. Judges have discretion to determine when a child becomes an adult. Judges frequently authorise the marriage of girls who have reached puberty. And an effort to set a minimum age for marriage was, in the Committee's words, 'successfully challenged in December 2014 by the highest-ranking religious leader, who declared being in favour of marriages involving girls as young as 9 years old.'
The Committee also recorded, in the same document, that 'judges rule on the basis of their personal interpretation of religion', and that the country's general reservation to the treaty 'provides for the precedence of sharia law over international treaties'. That reservation is what supplies the religious warrant, and the record names it — under the new rule a record cannot rely on the clause without identifying the provision.
Under the old rule this would have been excluded twice over: judicial discretion is state action, and a defeated bill is legislation. That exclusion is exactly the defect the rule change was made to correct — the more completely a state fused religion and government, the less of it this archive could see.
No individual is named. The Committee does not name the religious leader and neither does this archive; the record concerns an office and its effect, not a person, and asserts no offence by anyone. It is a finding by a UN treaty body, not a court: nobody was prosecuted. It gives no figure for how many marriages occurred, because the Committee gives none. And it describes 2016, not Saudi law today.
One thing this record is not: a finding about Islam or about Muslims. It is a finding about how one state's institutions operated, made by the treaty body that state reports to. The same rule reaches Israel's rabbinical courts and Vatican City, and describes what was already true of Ireland's Magdalene laundries and the residential schools of Canada and Australia.
The Data Bias page now reads '1 record relies on this rule', computed from the corpus. It said none this morning. That number is published so the effect of the change can be watched rather than taken on trust.

Effect on published figures

Institutional records rise from 144 to 145. Saudi Arabia goes from zero records to one, and the archive gains its first record sourced from a UN treaty body — a channel that reaches countries whose courts publish nothing.

July 28, 2026 records

What the money actually looks like: 500 roubles for a fortune, 424,000 by the third day

Two more Russian judgments. Both show something the headline figures hide — how a small opening payment becomes a large one, and what the perpetrator physically does to make it believable.

Pavlovsk, Voronezh Oblast, July 2020: a woman approached a stranger in a shop and offered to tell her fortune for 500 roubles. She then said a curse and the evil eye lay on her and sold her a pendant as protection. The next day she telephoned to say she had 'read the book of her family' and the curse was on every member of it. On the third day she asked for the victim's gold and silver to be handed over for ritual cleansing. Total taken: 92,022 roubles in jewellery, 311,830 in cash and 21,000 in goods — from an opening of 500.
The court recorded that the victim was susceptible to psychological influence and inclined to trust strangers. That is the court's own finding about how the offence was possible, and it is reproduced here as such — not as this archive's judgement about the woman who was defrauded.
Karpinsk, Sverdlovsk Oblast, May 2020: a woman approached strangers in a courtyard and a park and told them she had powers that could mend a mother's relationship with her son. She went into one victim's flat, asked for an egg and a towel, and performed several actions with them to create the appearance of a magical rite before asking for money. The court found the 22,000 roubles were handed over under psychological influence. Three victims, four years, suspended in full.
The egg detail is worth stating plainly. Passing an egg over a person with a cloth is a real Slavic folk practice for lifting the evil eye. The judgment finds this defendant created only the appearance of such a rite. The archive makes no claim about the practice itself or about anyone who performs it sincerely — the point is that a genuine tradition is what made the performance credible.
Neither defendant is named; Russian judgments are anonymised, so each is identified by case number and court.

Effect on published figures

Core records rise from 1,040 to 1,042 and Russia from 23 to 25. A search for sexual offences committed under ritual pretext returned nothing in this database, which is recorded as an honest empty rather than left as an untested assumption.

July 28, 2026 policy

A rule that was hiding abuse in exactly the states where religious authority is most total

The archive's inclusion test quietly assumed that religious and state authority can be told apart. Where they cannot, the test was excluding the very cases it exists to record. This is a defect of our own making, not one imposed by the world.

The test asks whether the coercion collapses once you subtract the perpetrator's religious authority. That question presupposes you can take religious authority away and still have state authority standing. Where a country's courts derive their warrant from religious law, you cannot: the judge is not a secular official who happens to be devout, the religious interpretation IS the judicial warrant, and subtracting it leaves no judge at all.
The working rule had been that state action falls outside the category. In a country with separated powers that is correct, and it is why an abusive police officer is not a record here. Applied to a fused state, the same rule excluded the cleric-judge because he was a judge, and a religiously-derived criminal ordinance because it was legislation. The effect was that the more completely a state fused religion and government, the less of it this archive recorded — which inverts the entire point of the archive.
From now on, where judicial, legislative or executive authority is constitutionally derived from religious law, exercising it counts as exercising religious authority, and the test runs against the doctrine rather than the office. Any record relying on this must name the constitutional or legal provision that supplies the religious warrant; it cannot use the rule without saying which one.
Four limits come with it and matter more than the rule itself. It does not reach ordinary crime — fusion shows religious authority was available, never that it was used. It is not about one religion: it reaches Israel's rabbinical courts and Vatican City, and it describes what was already true of Ireland's Magdalene laundries and the church-run residential schools of Canada and Australia, which this archive has always held. It does not touch persecution — a theocratic state prosecuting a religious minority is a record of persecution, never of authority abuse. And it does not improve anyone's evidence: a state-media report is still not a court judgment.
Two automatic checks now enforce this. One rejects any record using the rule without citing the provision behind it. The other rejects any attempt to apply it to a persecution record.
No record uses the rule yet. The change is published now, at the point it was made, rather than after it has quietly reshaped the figures — and the Data Bias page states the count, computed from the corpus, so you can watch it move.

Effect on published figures

No records were added or altered. What changed is which cases can be considered at all, in the places this archive has been least able to see.

July 28, 2026 records

Curse-removal fraud is an organised trade, and the judgments describe how it works

Two more Russian convictions read from the courts' own judgments. Together with those already added they show something the individual cases do not: this is not improvised opportunism but a repeated routine with assigned roles.

Sayanogorsk, Khakassia, July 2020: the court set out a three-way division of roles agreed in advance. One woman was to find an elderly woman in the street and offer help removing a curse and healing her. A second — the defendant — was to join the conversation to convince the victim that the help genuinely worked, walk her to where she would hand over her valuables, and keep pressing the need for treatment until she did. A third kept watch and was to distract anyone who approached. Convicted on two counts; an earlier suspended sentence was revoked, making four years in a penal colony.
Pavlovsk, Voronezh Oblast, February 2025: a woman offered a stranger in the street a rite to remove a curse, was invited into her home, and asked her to hand over all her savings so the rite could be performed over the money. She took the 240,000 roubles. Sentenced to three years, converted to compulsory labour.
The Pavlovsk case was prosecuted as theft rather than fraud, because the court found the money was taken once produced rather than handed over on the strength of the lie. The archive follows the convicting court's own label rather than reclassifying it. What matters for inclusion is that the rite was the means of getting into the house and at the savings — and on that the judgment is explicit, calling it an invented rite.
Read alongside the Kursk case added earlier — where one woman posed as the healer and another as a satisfied former client — the pattern is consistent across four regions and seven years: approach an older woman in public, name a curse, produce a second person to confirm the healer's power, and keep her talking until the money is out. The corroborator role appears in the courts' findings again and again.
No defendant in any of these cases is named, because Russian judgments are anonymised. Each is identified by case number and court instead.

Effect on published figures

Core records rise from 1,038 to 1,040 and Russia from 21 to 23. A refined search of the Russian judgment database returns 342 further judgments on this pattern, so this is a channel with substantial depth rather than a handful of cases.

July 28, 2026 records

Three curse-removal frauds read from the Russian courts' own judgments

Working through Russia's public judgment database directly. Three convictions added, one keyword match rejected — all read from the full text of the judgments rather than from reports about them.

Miass, Chelyabinsk Oblast, February 2020: a woman approached five strangers in the street in a single day, telling each either that she had a disease or that a 'death curse' had been laid on her, and that she was a healer who could lift it — on condition of unquestioning compliance with all her instructions. The court found she had no such powers. Convicted on five counts of fraud and sentenced to two years and six months in a penal colony, taken into custody in the courtroom.
Kursk, November 2019: two women worked a scripted routine on elderly women in the street. One posed as a healer offering to remove a curse; the other then approached and vouched that the healer had already removed a curse from her. The court found they agreed the division of roles in advance and knew throughout that they could not do what they promised. Both convicted on two counts and jailed for a year each. The court's finding that they set out to defraud 'some elderly woman', with no particular victim in mind, is recorded as the court's own.
Serdobsk, Penza Oblast, October 2025: a woman advertising online as 'the healer Lyudmila' charged a disabled woman 76,000 roubles for a rite to speed up delivery of her prosthetic leg. Three years, suspended — the record states she was released in the courtroom and did not go to prison.
One case that matched the search was rejected. A judgment from Cherkessk mentioned an invented healer called 'Magomet', but the case was an online fraud selling Apple phones that did not exist across nine counts; the healer was one pretext among many in a scheme that had nothing to do with belief. A keyword match is not a mechanism.
None of the three defendants is named, because Russian courts anonymise published judgments. Each record identifies its case by number and court instead. That is now the archive's stated rule, and it applies here for the first time at scale.

Effect on published figures

Core records rise from 1,036 to 1,038 and Russia from 19 to 21. All three rest on the full text of the judgment, which is stronger evidence than the press reporting the archive would otherwise have relied on.

July 28, 2026 policy

Court judgments that do not name the defendant are now admissible

A rule has been changed. Records were previously turned away when no source named the perpetrator. That excluded the primary court documents of every country whose legal system anonymises judgments — which is most of the civil-law world.

Russia, Germany, Poland and many other jurisdictions publish judgments with the defendant anonymised by law. In Russia the accused appears simply as 'FIO1'. Under the old practice, the court's own judgment — the strongest possible evidence, the actual primary document — was rejected, while a weaker English-language newspaper report naming someone would have been accepted. That biased the archive toward English-speaking, common-law countries. It was a defect, not a standard.
Such records are now admissible, identified by case number and court rather than by name. Five rules apply: identify by case number and court; never guess a name or use other sources to unmask someone the court chose to anonymise; state plainly in the record that the defendant is unnamed and why; treat a trading pseudonym as a persona and not as an identity; and victims remain unnamed regardless.
Nothing else changes. The requirement that the harm ran through spiritual authority, the requirement of an adjudicated outcome, and the appeal check all apply exactly as before. An anonymised judgment that fails those tests is still refused. This governs identification only.
This regularises what the archive was already doing rather than opening a new door: 106 existing records already state explicitly that the perpetrator is unnamed.
The first record under the new rule comes from the Serdobsk City Court in Penza Oblast, October 2025. A woman created a page on a Russian social network under the name 'the healer Lyudmila', advertising fortune-telling, love spells and the removal of curses she had no intention of performing. She charged a disabled woman 76,000 roubles for a 'magical rite' to speed up the delivery of her prosthetic leg. She was convicted of large-scale fraud and given three years, suspended — the record notes she was released in the courtroom and did not go to prison, so it should not be read as a custodial outcome.

Effect on published figures

Core records rise from 1,035 to 1,036. The more significant effect is prospective: primary court records from jurisdictions that anonymise are no longer excluded, which should reduce the archive's tilt toward English-language sources over time.

July 28, 2026 records

Searching court records in their own languages: first results from Japan

The archive has begun searching each country's court reporting in that country's own language rather than relying on English coverage. The first record from that work is a Japanese case whose judgment states the mechanism this archive exists to document more plainly than almost any other.

The head pastor of the Seishin Chuo Church in Kyoto was sentenced by the Kyoto District Court in February 2006 to twenty years' imprisonment for twenty-two sexual assaults on seven girls who were members of his church. He did not appeal, and the sentence became final three weeks later.
The court's own reasoning is the reason this record matters. It found he had abused the position of head pastor, 'whom the victims regarded as a being close to God', and called the offending of a depravity without parallel among sexual crimes. It further found the girls believed that disobeying him meant disobeying God, and that refusing him would mean falling into hell — and treated that belief as what made resistance impossible. That is a court identifying religious authority itself as the instrument of the crime.
The defendant is not named here. Japanese reference sources withhold the names of convicted individuals in cases of this kind, and no judgment text naming him was obtained. He is identified by his role and church, as this archive does elsewhere when a court record exists but a name does not. Two women who held office in the church were also reported arrested; no outcome for them was found and the record makes no claim about them.
The record is also explicit about the limits of its sourcing. The conviction became final because nobody appealed, not because a higher court affirmed it, and the record says so rather than implying appellate endorsement. The judgment text itself was not retrieved — the quoted findings come from two independent Japanese reference works that were cross-checked against each other on court, date, sentence, charges and number of victims.
On the wider search: Portuguese-language searching of Brazilian court and prosecutor reporting returned three strong cases in a row that were ALREADY in this archive, including a pastor sentenced to 136 years in Goiás and another to 37 years in Santa Catarina. Korean-language checking likewise found the major cases already held. Japan was the outlier — six records for a country of 125 million, all of them famous cases, with ordinary criminal convictions missing entirely.

Effect on published figures

Core records rise from 1,034 to 1,035. More importantly, the search method has changed: coverage is now being tested language by language rather than assumed from English-language results.

July 28, 2026 Correction

Every published UK figure was wrong by about a factor of three

A check of how the archive counts countries found the United Kingdom split across three separate buckets. Anyone who read a UK number from this site or its public data files got a number roughly a third of the real one.

The archive works out which country a record belongs to by reading the last part of its location text. That is an approximation and always has been. But it meant that records written 'London, UK' and records written 'Manchester, United Kingdom' were counted as two different countries — and 'Kidwelly, Wales' as a third.
The published effect: the United Kingdom appeared as 51 records under one name, 19 under another and 1 under a third. The correct figure is 71. This affected the public statistics file, the downloadable datasets and the country counts on the site.
Country names are now resolved to a single spelling before anything is counted, covering the UK's constituent countries and common variants for the United States, Ireland, the Netherlands, Czechia and Korea. The total number of countries in the archive accordingly drops from 171 to 168 — that is three duplicate entries merging, not three countries disappearing.
Two automatic checks now run on every build: one asserting that no record is filed under a known variant spelling, and one asserting that every record with a location can still be attributed to some country. The first would have caught this.
What has NOT changed: reading a country out of free-text location text is still an approximation, and the archive still says so. Records covering more than one jurisdiction, or historical entries whose borders have since moved, are not resolved by this fix and are not claimed to be.

Effect on published figures

No records were added, removed or altered. Only the counting was wrong. If you have previously downloaded the dataset or quoted a per-country figure from this archive, the UK figure in particular should be taken again.

July 28, 2026 records

A South African killing where the witchcraft accusation was itself a criminal charge

One record added from South Africa's Truth and Reconciliation archive, and three long-standing research leads worked to a conclusion — two of which end in 'not enough', which is recorded rather than hidden.

Michael Morudi Phasha and Daniel Lisufi Phasha were convicted in the Lydenburg Regional Court in June 1995 of the murder of Abram Madibeng Phasha, killed at Driekop in February 1993, and each sentenced to an effective thirteen years. What makes this record unusual is the second conviction: they were also convicted of imputing witchcraft to the man they killed, a distinct offence under South Africa's Witchcraft Suppression Act of 1957. The accusation was not merely reported background — it was a charge, tried and proven.
This is one of the archive's inverse records, and it is filed as such. The dead man was not under his killers' spiritual authority; he was killed because witchcraft was attributed to HIM. Records of that shape sit in a separate category and should never be read as a religious figure harming followers — the direction of the belief is reversed.
The record carries a second official account without choosing between them. In 2000 the Truth and Reconciliation Commission granted both men amnesty, finding the killing politically associated: perceived witchdoctors, it held, were regarded as political enemies because they were seen as aligned with the local Chief. So the state's own archive contains both a witchcraft-accusation conviction and a political-killing finding about the same deaths. Both appear here. The record also notes that amnesty extinguished liability but did not overturn the convictions — it is not an appeal outcome and is not counted as one.
Three leads were worked and NOT added. Isaac Carpin, an Adventist pastor in Martinique sentenced to 14 years in 2021 for rapes and sexual assaults on five girls, is a real conviction — but the trial was held behind closed doors, so no public source establishes how his pastoral role gave him access, and this archive requires that link rather than assuming it from the job title.
A Trinidad case with the starkest mechanism in the whole review — a self-described spiritual healer who told a sixteen-year-old that her dead father's spirit had to be exorcised through sexual penetration, staging the first assault on a grave — is also not included. As of now only charges and bail have been reported, ten months on. No verdict, no record. Nothing about the strength of an allegation substitutes for an adjudicated outcome.
A note on method: the source citation for the South African decision was wrong in the archive's own internal research notes, pointing to an unrelated amnesty case. It was caught by reading the cited document instead of trusting the note, and the correct decision was located and read before anything was written down.

Effect on published figures

Historical records rise from 8 to 9, and accusation-violence records from 10 to 11. Two researched leads were closed as insufficient and one as unadjudicated.

July 28, 2026 records

One case added from Brooklyn, and three well-known ones examined and turned down

A search of prosecutions in Brooklyn's Orthodox community produced one record and three refusals. The refusals are the more useful part, because two of them carry very long sentences and would have looked impressive here.

Israel Weingarten, convicted by a federal jury in 2009 of transporting his daughter across international borders to abuse her, sentenced to thirty years. The abuse itself began when she was nine or ten and ran more than six years, but it happened in Belgium and Israel, so the American charges are for the travel rather than for the abuse — the record says so rather than letting the sentence imply otherwise.
Abuse by a parent is not normally within this archive's scope. Weingarten is here because the appeals court's own account of the trial evidence describes something beyond the family: he told his daughter she would not be believed 'because of the respect in which he was held in the Satmar community'; he made her record a conversation designed to make her look promiscuous, which he said he would play to community rabbis to support his denials; and the allegations went before a rabbinical court where she testified against him. The standing he held in the community was the instrument for defeating disclosure.
The full appeal history is recorded because it changes the picture. One of the five counts was reversed in 2011, so the record is marked as modified rather than affirmed, even though he was resentenced to the same thirty years, that sentence was affirmed in 2013, and the Supreme Court declined to take the case in 2018. The record also makes no finding about the rabbis or the rabbinical court, several of whom the opinion says acted to get the daughter out of the house and into a school abroad.
Emanuel Yegutkin, a Brooklyn yeshiva principal sentenced to 55 years, is NOT included. The reporting is explicit that the three brothers he abused were not his students — he reached them by befriending their father and became a regular guest in their home. His title was religious; his access was social, and this archive tracks the second, not the first.
Jacob Daskal, founder of the Borough Park Shomrim, sentenced to 210 months in federal prison, is NOT included either. Prosecutors say the girl came into his care because of his position with the Shomrim — but that is a neighbourhood security patrol working with the police, not a religious office, and nothing in the record shows doctrine or spiritual claim doing the work.
Avrohom Mondrowitz, probably the best-known name in this area, is NOT included. He was indicted in 1985 and fled to Israel, but Israel's Supreme Court refused extradition in 2010 and a separate case there was closed for lack of evidence. He has never been convicted of anything. This archive records convictions and official findings, not accusations, however old or well known.

Effect on published figures

Core records rise from 1,033 to 1,034. Three widely-reported cases were researched to the court record and excluded, which is the ordinary result of applying the test rather than a failure of the search.

July 28, 2026 records

A Brooklyn conviction, and six records that were filtering into the wrong bucket

One new record from the New York courts, and a repair to how six existing Jewish-institution records were being categorised on the site's filters.

Yona Weinberg, Brooklyn: convicted after trial in September 2009 of seven counts of sexual abuse in the second degree and two of endangering the welfare of a child. The Appellate Division affirmed the conviction in full in July 2010, calling the evidence of guilt overwhelming. He served about thirteen months and was designated a level 3 sex offender.
The Weinberg record deliberately separates the two victims rather than treating them alike. One was a boy he taught bar mitzvah lessons to at a synagogue — that is authority of the kind this archive tracks. The other was a client of his separate work as a licensed social worker. Professional and custodial access is NOT what this archive counts, and the record says so plainly; the second count appears because it is part of the same judgment, not as a second example of the pattern.
The record also marks where each fact comes from. The conviction, the counts, the court, the judge and the affirmance are taken from the published appellate opinion. That opinion does not describe how Weinberg reached either boy, so the tutoring role is attributed to news reporting instead, and the record says which is which rather than letting the court lend its authority to a fact it never stated.
Six existing records under Judaism carried no tradition-detail value, meaning they were invisible to that filter on the site even though they were in the archive. They have been assigned: the Bratslav and Lev Tahor matters to Hasidic and Haredi communities, the Knesset hearing and the Marc Gafni record to the general Jewish-institution category, and two older Brooklyn and Woodridge records to Hasidic and Haredi. No record's substance changed — only whether it can be found.

Effect on published figures

Core records rise by one. Judaism now stands at 36 records, all of which now appear correctly under the tradition-detail filter.

July 28, 2026 records

Pelletstown and Kilrush, and one Irish institution examined and left out

Two more institutions from Ireland's Mother and Baby Homes Commission, including the one with the largest recorded death toll of any it examined. A third was read in full and not added, and the reason is given.

Pelletstown, on the Navan Road in Dublin, run by the Daughters of Charity of St Vincent de Paul: the Commission identified 3,615 deaths of infants and children associated with the home — more than at any other institution it examined, and more than three times the figure for Sean Ross. It found a state death record for 3,511 of them. Deaths were concentrated between 1920 and 1942, peaking at 200 in 1920 and again at 188 in 1936.
For Pelletstown the archive records the government department's own table of numbers resident against numbers dead, year by year through the 1920s: 96 deaths among 259 children in 1924, 119 among 240 the following year, and so on through 1930. It also records what happened when the death rate was raised at the time. Told in 1925 that the proportion of children dying was 'very heavy', officials replied that this was to be expected when infants were housed together, and the Board of Guardians' representative blamed children admitted from the street, who he said 'come into the world severely handicapped and are not normal at all'.
Pelletstown's record states a difference rather than hiding it. Unlike the Sacred Hearts and Bon Secours homes already here, Pelletstown was owned by the Dublin Board of Guardians, its Sisters were local government employees, and the Commission calls it an integral part of the public assistance system. The congregation ran it, but inside a public structure. Readers comparing institutions should not treat the arrangements as identical.
Kilrush, the County Clare Nursery, run by the Sisters of Mercy from 1922 to 1928: this record turns on a response rather than a toll, because the Commission says plainly that the records do not establish how many children died there. In March 1927 the medical officer reported the death rate as 'appalling'. Three months later the department wrote to the board objecting that the nursery's allowances of bread, potatoes and butter were 'substantially in excess' of the departmental scale and that it saw 'no adequate reason' for them. The archive also notes the Commission's observation that the board minutes contain continuous requisitions for coffins in different sizes.
Regina Coeli in Dublin was read in full and deliberately NOT added, though 734 child deaths are recorded there. It was run by the Legion of Mary, whose religious mission the Commission found central to its operation — but its policy was to keep mothers and children together, the opposite of the other homes, and a medical inspector's finding was that the voluntary effort, 'beyond doubt laudable', was unable to cope. That is a failure of capacity and funding, not authority used to make harm unrefusable, and this archive only holds the latter. The Commission also judged that confirming those deaths against the state register was not feasible, so the figure does not stand on the same footing as the others here.

Effect on published figures

Institutional records rise from 142 to 144. Five of the Commission's eighteen institutional chapters have now been examined, four added and one excluded on the record.

July 28, 2026 records

Tuam and Sean Ross: two more Irish institutions given their own records

Ireland's Mother and Baby Homes Commission examined each institution in a separate chapter. This archive has been adding them one at a time, using the Commission's own figures. Two more, including the home whose burial ground prompted the inquiry.

Tuam, County Galway, run by the Sisters of Bon Secours: the Commission identified 978 deaths of children who had been in the home — 802 in the home itself, 80 at its predecessor at Glenamaddy, 80 in a Galway hospital and 8 elsewhere — and saw the state's own death records for 972 of them.
On conditions at Tuam, the Commission reproduced the figures a government inspector recorded year by year: 54 deaths against 159 births and admissions in the year to March 1943, then 25%, 23% and 27% in the three years following. The rate then rose again, to about 32% in the half-year to September 1946. The inspector's list of what the children died of was said to be attached to her report; it was not in the file the Commission received.
Sean Ross Abbey, County Tipperary, run by the Congregation of the Sacred Hearts of Jesus and Mary: 1,090 children who were born in or admitted to the home died in infancy or early childhood — more than at any other institution recorded here — with 97% of those deaths confirmed against the state register. Most occurred between 1932 and 1947.
One line in the Sean Ross chapter carries the finding about official knowledge. For the year to March 1937 the government department's own report recorded 120 births and 60 infant deaths at the home. In the Commission's words, 'no comment was made about this very high death rate.'
Both records state what they do not establish. No one was prosecuted over either institution and no individual is named. The mortality figures are counts of deaths among children in the homes' care, not findings that any particular death was unlawful. Neither record makes any claim about where the children were buried: the Commission dealt with burials in separate reports, and the Tuam record additionally notes a finding the Commission did NOT make, so it cannot be read as asserting it.

Effect on published figures

Institutional records rise from 140 to 142 and sources from 2,271 to 2,273. The archive now holds three of the Commission's institutional chapters alongside its national finding of roughly 9,000 child deaths across eighteen institutions. Fourteen chapters remain unextracted.

July 28, 2026 schema

Dates now say how much of themselves is actually known

Roughly a quarter of the dates in this archive were never full dates. A proceeding known only to a year had to be written as the first of January, and one known only to a month as the first of that month — which made them indistinguishable from proceedings that genuinely fell on those days, and from invented dates. Every date now carries a statement of its own precision.

Each proceeding now records whether its date is established to the day, to the month, to the year, or not at all. Of 1,423 proceedings: 937 are known to the day, 186 to the month, 164 only to the year, and 136 have no established date.
The 164 year-only entries are the reason this was needed. They were stored as 1 January and read as though someone had looked up an exact date — when what the source actually gave was a year. Some are genuinely year-scale: a panic that ran across a year, or proceedings still ongoing.
Anyone drawing a timeline or counting dates from this data should now filter on the new field. Treating the stored day of a year-only entry as real puts a false spike of 164 proceedings on 1 January — larger than any genuine single-day cluster in the archive.
Two rules are enforced when the archive is rebuilt. A proceeding claiming to be known to the day, while carrying a 1 January date, must record how that date was established — so a placeholder cannot be quietly upgraded into a fact. And a proceeding cannot claim any precision at all if it has no date.
The field appears in the published data files and the downloadable spreadsheet of proceedings, so the distinction survives outside this website.

Effect on published figures

No record's content changed and no case was added or removed. What changed is that a reader can now tell a date the source gave from a date the format demanded. This is the same correction made earlier to how the archive records whether a conviction survived appeal: in both cases a field was silently carrying two different meanings, and in both cases the fix was to make the archive say which one it means.

July 28, 2026 corrections

Six duplicate records removed, and one record added — the public count falls by five

Six cases were in this archive twice, each under two different web addresses. They have been reduced to one record each, so the number of cases shown publicly drops. Nothing was lost: every fact in a removed record was already present, and better sourced, in the one that remains.

Five of the six shared a common origin. An early pass created placeholder entries from an index page, giving each a single citation to a Wikipedia article and a proceeding date of 1 January — a stand-in for a date that was never established. Later passes researched the same cases properly and wrote them up under clearer addresses, but the placeholders were never withdrawn. The duplicates were in Portugal, Latvia, Nicaragua, Mexico and Mississippi.
In two cases the archive kept the better record rather than the one first proposed for keeping. The Mexican pair was resolved in favour of the entry carrying the actual sentencing date and a full article citation, over one carrying the 1 January placeholder and a link to a news site's front page. A Zimbabwean pair had already been merged earlier in the day, in the opposite direction to what was later proposed; acting on the proposal would have deleted the surviving record and left the case unrecorded.
A sixth duplicate was found while checking the others: a Mexican pastor's case existed twice, and the version kept is the one recording both of his convictions, in 2021 and 2023, rather than only the later one.
Two removals mattered for more than tidiness. A Nicaraguan duplicate claimed in its own notes to be a separate proceeding; it was not, and the record that remains carries a caution about irregularities in the trial that the duplicate omitted. A Mississippi duplicate described the outcome as a conviction without recording that the jury could not agree on the count concerning a second complainant, which the surviving record preserves.
Added: Andrew Roy, convicted at Sheffield Crown Court in 2017 of four counts of indecent assault against a boy he abused from the ages of seven to ten, in the boy's home and at a Sunday school camp he ran. The boy did not tell anyone as a child because Roy told him it would upset his father and that no one in the church would believe him. The record attributes the offending to an evangelical organisation connected to the United Reformed Church, which is how the inquiry describes it, and not to that church itself.

Effect on published figures

Public records fall from 1,304 to 1,299 — six removed, one added. Records recording a conviction fall from 1,038 to 1,033 for the same reason. Anyone who has counted cases from this archive before today should re-derive the figure: the fall reflects six cases that were counted twice, not any case being withdrawn.

July 28, 2026 records

Two Tanzanian ritual killings added, after a rejection that got the rule backwards

An earlier version of this entry announced one record and explained why a second, very similar case had been left out. That explanation was wrong, and the second case is now included. The error is described here rather than quietly fixed, because the reasoning is the part of this archive that can be checked.

Tanzania, 2026: the Court of Appeal upheld the conviction and death sentence of Emmanuel Safari for the murder of his one-and-a-half-year-old stepson. The court's account is that Safari, 'having been instructed by a local practitioner of witchcraft that ritual sacrifice would bring him prosperity, turned his attention to his stepson'.
Tanzania, 2026: Joseph Muhuila was sentenced to death for the murder of his four-year-old son Timothy, killed for ritual purposes in April 2025.
The second case was initially rejected on the ground that a father killing his own child is a family matter rather than one of religious authority. That reading of this archive's own rule was wrong. The rule excludes a devout parent whose crime does not depend on the belief; it expressly includes a parent whose act is genuinely driven by doctrine. A killing carried out as a ritual sacrifice is driven by doctrine by definition — remove the belief and there is no killing.
The archive's existing records already settled the point. It holds a South African case in which four relatives were convicted of a ritual murder with no ritual specialist named, and an Oregon case in which parents were convicted over the death of their own child because doctrine dictated what they did. The rejection contradicted both.
The earlier entry also presented the two Tanzanian cases as an instructive contrast — one in, one out, differing only in whether a practitioner gave the instruction. That contrast was the error itself, and has been removed from the published record rather than left to circulate.

Effect on published figures

Records recording a conviction rise from 1,035 to 1,037. Tanzania holds three records where it previously held one. Anyone who read the earlier version of this entry should disregard its account of why a case was excluded: the case is now included, and the rule it cited says the opposite of what the entry claimed.

July 28, 2026 records

Bessborough: 923 child deaths, and three of every four infants dying in a single year

Ireland's Mother and Baby Homes Commission was already represented here by its national finding. Its report also examines each institution separately, and the largest of them now has its own record, taken from the Commission's own published figures.

Ireland, 2021: the Commission established that 923 children born in or associated with Bessborough — the Cork home run by the Congregation of the Sacred Hearts of Jesus and Mary from 1922 to 1998 — died in infancy or early childhood, and confirmed 99% of those deaths against the state's own register of deaths.
Almost 11% of all the child deaths across the home's 76 years happened in 1943 alone. In the Commission's own words: 'for every 100 babies born in, or admitted to, Bessborough that year, 75 subsequently died in infancy.' Among infants born to private patients the rate that year reached 80%. Mortality fell to just over 12% by 1946 and to around 2% from 1961 until the home closed.
The Commission also recorded 31 maternal deaths associated with the institution. It found the state's death records for 29 of them, and burial records for 12.
The record says what it does not establish. No individual was prosecuted over Bessborough and none is named. The mortality figures are rates of death among infants in the home's care, not a finding that any particular death was unlawful. The Commission's separate finding that the congregation kept no burial records for children who died in its homes, despite its own church law requiring them, sits in a chapter this archive could not retrieve, and is noted as unquoted rather than paraphrased.

Effect on published figures

Institutional records rise from 139 to 140 and sources from 2,271 to 2,273. The archive's existing national-level record for the same Commission — roughly 9,000 child deaths across eighteen institutions — remains separate; this covers the single institution with the largest documented death toll.

July 28, 2026 records

A pastor jailed for assaults during 'deliverance sessions', and an account the inquiry recorded but no court has tested

Two records from the same UK inquiry, deliberately filed at different levels of certainty. One is a criminal conviction. The other is an account given to a support organisation and relayed to the inquiry, with no charge and no named person, and it is marked as such.

United Kingdom, 2017: John Wilson, pastor of the Liberty Pentecostal Church in Keighley, was jailed for 21 years at Bradford Crown Court for a series of assaults on six women running from the mid-1980s to 2010. The women were assaulted during one-to-one 'deliverance sessions' which he said God had commanded him to perform to rid them of evil spirits. His wife received a suspended sentence for aiding and abetting, the court having heard she held down one of the victims; his church assistant was also convicted.
That record states the acquittal alongside the convictions. Wilson was charged with rape and was acquitted of it: the judge concluded he could not be properly or safely convicted on the relevant evidence. The prosecution opened on 31 counts across seven complainants; the convictions concern six women, and the record follows the convicted counts rather than the charges as opened.
The inquiry's own finding is what establishes the mechanism rather than merely the offence: deliverance is a recognised ministry in the denomination the church was associated with, and the inquiry stated it 'should not involve the removal of clothing or any intimate touching of body parts'. That is independent confirmation the rite was being used as cover, not followed.
Also added, at a lower level of certainty: the inquiry recorded an account of a teenage girl abused by a baba, a Sikh holy man, engaged by her family when she became unwell. He told her Guru Nanak had come to him in dreams and her illness would pass, gave her holy water, said he would massage her stomach to make the holy water take effect, and assaulted her while reciting a term of reverence throughout.
That second record is filed as context, not as an established case. There is no conviction, no charge, and no named person on either side. It is kept because the mechanism is unusually complete — belief supplies the access, the compliance and the framing of the act — and because excluding a documented account would distort a tradition this archive covers thinly. The inquiry gives comparable accounts involving Muslim, Pentecostal and Hindu figures in the same passage, and the record says so, because it should not be read as singling out one religion.

Effect on published figures

Records recording a conviction rise from 1,035 to 1,036; context records from 108 to 109; sources from 2,267 to 2,271. One further candidate from the same report was examined and left out: a Buddhist order's founder was recorded as having had relationships with up to 24 adult followers, some of whom said their consent was compromised by their respect for him as their teacher, but the inquiry reported that as their own account rather than as a finding, no charge ever followed, and he died in 2018.

July 28, 2026 records

A first-of-its-kind Vatican conviction, and a Madrid court on a priest's 'logical ascendancy' over altar servers

Two records added from court reporting in Italian, Polish and Spanish. One is the first conviction in history for sexual abuse committed inside Vatican City. Several other jurisdictions were searched and returned nothing new, because the cases they surfaced were already here.

Vatican City, 2024: the Vatican's appeal court overturned a 2021 acquittal and convicted Gabriele Martinelli, who as senior coordinator of student activities at the St Pius X pre-seminary had used violence and threats to compel a younger residential student — around fifteen at the time — into sexual acts between 2008 and 2009. He was sentenced to two years and six months and fined €1,000. Reporting describes this as the first conviction in history for such offences committed within Vatican territory. The record carries the earlier acquittal alongside the conviction rather than omitting it.
Spain, 2024: the Provincial Court of Madrid sentenced a priest of the Santiago Apóstol parish in Torrejón de Ardoz to ten years for the continuous sexual abuse of two altar servers, aged twelve and fourteen when it ended in 2019. The court found he 'took advantage of' the ministry he exercised and the relationship it gave him with the boys, holding that he 'possessed logical ascendancy' over them, and that the abuse happened weekly in the sacristy where his office put him alone with them. The conviction remains open to appeal and is recorded as not final.
Searches in Polish, Italian and Portuguese returned strong cases that were already in this archive: two Polish priests whose convictions are final, an Italian priest whose conviction was made final by the Court of Cassation, another sentenced to ten years for abusing a twelve-year-old altar boy, a Portuguese priest convicted on eighteen counts, and an Azores case. Finding the same cases twice is the ordinary result of a well-worked source, and is recorded rather than presented as new.

Effect on published figures

Records recording a conviction rise from 1,033 to 1,035 and sources from 2,265 to 2,267. The Philippines was also worked as an apparent gap and produced nothing: its coverage was undercounted by a flaw in how this archive's own country tallies were computed, and the leading public database of accused Philippine clergy turned out to hold 95 entries but only five criminal convictions, four of them tried in United States courts.

July 28, 2026 records

Two findings added from the UK child protection inquiry, on rules that stopped abuse being reported

The 2021 statutory inquiry into child protection in religious organisations examined 38 settings across England and Wales. This archive held findings from only four of them. Two more are added, both concerning a religious rule or practice that operated to keep abuse away from the police — which is the test for inclusion here. Several other settings named in the same report were checked and left out, for the same reason the Quakers were.

Charedi Jewish communities: the inquiry found that the concept of mesirah — that for one Jew to report another to a non-Rabbinic authority is forbidden, believed to be required under Halacha — may prevent abuse being reported outside the community. A witness described it being used 'as a weight and leverage against victims and survivors in reporting their abuse to the authorities', and the inquiry recorded a senior rabbi using the term on camera in a 2013 documentary to counsel a survivor against going to the police. A survivor who did consider reporting was offered £5,000 and told this was the 'only route', because going to the police would mark her a 'Moiser' and see her shunned in a community that is also her family, employment and social world.
That record carries the inquiry's own qualifiers rather than stripping them. The finding is about 'some in Jewish Charedi communities', not the communities as a whole. The inquiry expressly records why the tradition exists — centuries of persecution and fear of unfair treatment by secular authorities. And it records the contrary rabbinic position, that mesirah does not apply where a person is causing harm to others such as child sexual abuse, alongside its own assessment that the rabbinate's 2013 statement was not sufficiently clear.
The Jesus Fellowship Church: the inquiry found it was 'impossible' for children in its communes to disclose abuse, because they were not believed, or 'the abuser would have been spoken to, prayed over, possibly moved to another house with other children'. Those who tried to leave were 'spoken ill of, cursed' and 'prayed over very, very heavily by groups of men, rebuked in public, shouted at, screamed at', and children were taught that awful things would happen to them if they left. Six people were later convicted of offences relating to eleven children after the church invited disclosures in 2013 and police opened an investigation.
This archive already held three of those individual convictions and the church's redress scheme report. What the new record adds is the inquiry's account of how the church handled abuse internally, which is what the conviction records do not cover.
Other settings named in the same report were checked and left out. Buddhist and Sikh organisations, the Salvation Army and the Baptist Union appear in it substantially, but what the inquiry says about them concerns policies, record-keeping and audit arrangements rather than any religious rule or practice that enabled abuse or suppressed its reporting. Being examined by an inquiry is not the same as being the subject of a finding, and that distinction decides what enters this archive.

Effect on published figures

Institutional records rise from 137 to 139 and sources from 2,263 to 2,265; the count of individual convictions is unchanged. Both new records are inquiry findings, not criminal findings, and neither names an individual as a perpetrator.

July 28, 2026 corrections

A bug was hiding appellate outcomes this archive had already recorded

The code that reads each record's appellate position was looking in the wrong place. It read the narrative account of a proceeding but not the field where this archive actually records what an appeal decided — so scores of records were marked unverified when their own text already said the conviction had been upheld. Fixing it moved 116 records out of the unverified column without any new research. A separate correction reverses a change made in error yesterday.

The appellate check now reads the disposition field of every proceeding, not only the narrative. Records saying 'jury conviction; largely affirmed on appeal' or 'conviction and death sentence; appeals exhausted' had that written down all along; nothing in the code was reading it. Records with an established appellate position rise from 283 to 336 of 1,035 convictions, and the unverified count falls from 727 to 669. This is recovered work, not new research: no source was consulted that this archive had not already cited.
A related ordering error is fixed. Where a record narrates a full appellate history, the code was stopping at the middle of it. One case had a sentence overturned and a retrial ordered, and then the retrial verdict upheld on final appeal; another had a conviction twice reversed by an appeal court and then reinstated by the state supreme court. Both were being recorded as sentence changes when the last appellate word had restored the conviction outright. A final affirmance now takes precedence, and both are recorded as upheld.
The distinction that matters is preserved: a sentence altered on appeal with guilt left intact is still recorded as a sentence change, not as an upheld conviction and not as an overturned one.
Reversed a change made yesterday in error. A record was altered to move its location from Texas to Alabama because the prosecution took place in Alabama. That was wrong: this archive records where events happened, which is not always where the case was tried, and the record already named the Alabama court correctly. The location is restored to Texas. The verified case number and the note that the case is still open both stand.
Wrongful-conviction records remain excluded from all of this, as they must be. Their appellate histories often contain the word 'affirmed' describing appeals that preceded an exoneration, and reading that as a standing conviction would invert what the record documents.

Effect on published figures

No records were added or removed and no evidence changed. The published figure for verified appellate positions rises to 336 of 1,035 convictions (32%), with 669 still unverified. That the correction was possible at all is the argument for having created the field: the information was in the archive, and nothing surfaced it until the field forced the question.

July 28, 2026 corrections

Federal appellate check complete: four recorded convictions were not final, and two named the wrong place

Every record in this archive resting on a United States federal prosecution has now had its docket checked one at a time. Twenty were checked in detail. Four turned out not to be final — one of them recording a conviction whose sentencing has not yet taken place — and two named the wrong city and state. None of that was visible before the check, and all of it is now stated on the records themselves.

A Colorado record described a conviction on fifteen federal counts. The jury verdict is real, but sentencing has not happened: the hearing set for April 2026 was vacated and reset to 18 August 2026. The record now says so and is marked for revisiting after that date.
A record of an online Quran instructor's conviction gave the place as Euless, Texas. The prosecution was brought in the Northern District of Alabama, which the record's own account of the proceedings already said. The place and map position are corrected to Birmingham. The federal case remains open, so the record is also marked as not final.
A Maryland record of a guilty plea entered in April 2026, with sentencing still to come, is now marked as not final. A candidate federal docket was examined but its entries were blank, so the defendant could not be independently confirmed — and no docket number is asserted on the record rather than one being cited on a partial match.
Five records were searched for appeals and none was found: a Buffalo case involving a retired judge, an Oklahoma case, the United Nation of Islam sentencings in Kansas, a Tacoma case, and a federal prison chaplain's case in California. Each now carries the docket number that was checked. Two of these are only findable by docket number, because the case is captioned by a defendant's name rather than by the organisation the record is about.
One record was deliberately left unverified. A 2015 Pennsylvania state conviction has a separate federal case against the same person in New Jersey, and a later habeas petition. Neither is an appeal against the Pennsylvania conviction, so neither is treated as its appellate history, and the state appellate record has not been searched. Saying that is more useful than implying the question was settled.
Also noted on one record: its stated verdict date falls after the date its federal docket closed, which cannot both be right. The discrepancy is recorded rather than resolved by silently changing the date to fit the docket.

Effect on published figures

Across the twenty federal records checked in detail, four were not final and two named the wrong place — a defect rate of roughly thirty per cent in records that read as settled convictions, none of it detectable without opening the dockets. This is the clearest evidence so far for why this archive publishes what it has not verified. Records with an established appellate position now stand at 283 of 1,035 convictions, and 25 convictions are marked as still running. Anyone using this dataset to count convictions should read the appellate field alongside the conviction field.

July 28, 2026 corrections

A recorded conviction is under appeal, and a record named the wrong court and state

Checking federal dockets for the appellate backlog turned up two things more important than the backlog itself: one record asserts a conviction that is currently on appeal and did not say so, and another named the wrong court, the wrong city and the wrong state. Both are corrected. A defect in how the build handled legacy records was also found and fixed.

The OneTaste record is now marked as under appeal. The convictions of Nicole Daedone and Rachel Cherwitz are before the United States Court of Appeals for the Second Circuit in three docketed appeals, none of them decided. This archive's rule is that a conviction which does not stand does not qualify, so a record asserting a conviction must say when that conviction is not final. The record now says so, and is flagged for review when the Second Circuit rules. Three separate proceedings that are commonly confused with this appeal are named in the record so they are not mistaken for it.
The Richard Densmore record gave the court as the Eastern District of Pennsylvania and the place as Philadelphia, with Philadelphia coordinates. The prosecution was in fact brought in the Western District of Michigan and the sentencing took place at Lansing; the defendant was of Kaleva, Michigan. This is confirmed both by the federal docket and by the Justice Department's own Western District of Michigan announcement, which has been added as a source. The court, place and coordinates are corrected and the error is stated in the record rather than quietly repaired. The record's web address still contains 'philadelphia-pa' and has deliberately been left unchanged so existing links keep working.
Three further records — a former priest in Ohio, a pastor in Virginia, and the Lev Tahor prosecution in New York — were searched for appeals and none was found. They are now marked as searched rather than left silent, and each carries the docket number that was checked. Records with an established appellate position rise to 278 of 1,035 convictions.
Fixed a defect in how the site is built. The 71 oldest records can only be refined through a separate overrides file, and the build was not reading appellate status from it — so a researcher could establish that one of those convictions had been upheld or was under appeal, record it, and have the value silently discarded. Those records could never have been marked as anything other than unverified. The passthrough is added and a test now guards it.

Effect on published figures

No records added or removed. Anyone citing the OneTaste record should note the convictions are on appeal and not final. Anyone who used the Densmore record's jurisdiction or map position should re-read it: it was Michigan, not Pennsylvania. The wrong-jurisdiction error was found because a docket number was checked against the record, which is the only way this class of error surfaces — a coordinate in the correct hemisphere and a plausible court name pass every automated check the corpus has.

July 28, 2026 records

Three convictions added, including a High Court affirmance; four candidates rejected on the criterion

A continuation run working court reporting in India, Brazil and Indonesia. Three records added and four rejected — one because reading the actual court file showed it was not an abuse case at all, two because an inquiry naming an institution is not the same as an inquiry making a finding against it, and one on the family-nexus rule.

India, 2026: the Madras High Court (Madurai Bench) affirmed the life sentence of Joseph Raja, pastor of a church at Rajapalayam, for the aggravated sexual assault of a seventeen-year-old girl with an intellectual disability. She and her mother regularly attended his church and sometimes stayed overnight there; on the night of the assault her mother left her in his care. The Court held the evidence 'incontrovertibly establishes the guilt of the appellant' and rejected the argument that her disability undermined her testimony. This is one of a minority of records with a verified appellate position.
India, 2022: Pentecostal pastor Jose Prakash was sentenced to life by the Manjeri fast-track POCSO court for raping a thirteen-year-old girl after telling her family she was possessed by evil spirits and needed 'special healing prayers' — the claim that put her alone with him. No appellate disposition has been located, and the record says so rather than assuming the conviction is final.
Brazil: Vanderlei Antônio de Oliveira was sentenced to 136 years at Anápolis, Goiás, for abusing congregants he told he could 'incorporate angels' to solve their problems; the demands were issued in the angels' purported voices during prayer sessions, and he filmed the assaults to coerce continued participation in what he called 'spiritual campaigns'. His wife received 95 years for witnessing and concealing the offences. Recorded as not final — both retain appeal rights.
Rejected — an Episcopal Diocese of New Jersey case flagged as the most likely US institutional record. Reading the court file rather than the docket summary showed it is an employment-discrimination suit brought BY a priest AGAINST his diocese over race discrimination and wrongful discharge, and it was dismissed. The plaintiff is the clergyman, not someone under another's spiritual authority.
Rejected — the Salvation Army and the Baptist Union of Great Britain as institutional records from the 2021 IICSA report into religious organisations. The full report was retrieved and searched: every finding about both bodies concerns governance, record-keeping and auditing, and several are favourable. Neither has anything resembling the finding that put the Jehovah's Witnesses record in this archive from the same report, namely that elders discouraged reporting and substituted internal discipline. Being examined by an inquiry is not the same as being the subject of one.
Rejected — a Nigerian pastor sentenced to 49 years in Akwa Ibom. The victim was his biological daughter and nothing in the record shows religious authority was the means of the abuse rather than the fact of being her father. A parent who invokes doctrine to compel a child is in scope for this archive; a parent who does not is out.

Effect on published figures

Core records 1,030 to 1,033 and sources 2,256 to 2,262. Records with an established appellate position now stand at 274 of 1,035 convictions. Four rejections are published here rather than left silent, because the reasoning is the part that can be checked: a defect in it is easier to spot than a missing record.

July 28, 2026 records

Seven institutional records from the New Zealand Royal Commission; Anglican and Methodist gain their own categories

New Zealand's Royal Commission of Inquiry into Abuse in Care examined eight faith groups. This archive had extracted findings for one of them. Reading the relevant chapter in full produced six further records — the Plymouth Brethren record published earlier today was the seventh — each resting on the Inquiry's own findings about the institution named. Every paragraph cited was read and quoted from the report itself before drafting.

Gloriavale: the Inquiry found victim-blaming 'particularly evident' there, with the founder teaching that women could avoid sexual assault by their dress and conduct, and current leader Howard Temple accepting that young women may have felt assault was their fault as a result. Until 2012 the community had 'little to no dealings with police', handling allegations internally under its doctrinal text so as to lead 'transgressors to a place of repentance, and victims to a place of forgiveness'.
Salvation Army New Zealand: 238 claims of historical abuse and neglect confirmed as at 1 August 2020, with the Salvation Army accepting the abuse in its care was wide-ranging. Women in its Bethany Homes were made to feel shamed as unwed mothers and pressured to adopt their children while denied information and medical support.
Methodist Church of New Zealand: the Church accepted that its Conference 'failed to provide sufficient monitoring, oversight and safeguarding' of three children's homes and 'that these failures enabled abuse and neglect to occur', and took full responsibility for those abused in its care and at Wesley College Pukekohe.
Catholic Church in New Zealand: 26 reports of alleged abuse relating to nine Society of Mary members at St Patrick's College Silverstream between 1951 and 1985, and acknowledgement of the harm at Marylands School and the Hebron Trust, where many of those in care were disabled boys, 'street kids', and tamariki and rangatahi Māori. The Inquiry identified the teaching that priests undergo an 'ontological change' at ordination as a source of the clerical power differential.
Anglican Church in Aotearoa New Zealand and Polynesia: acknowledgement that children were abused in its care, that members of the church covered up abuse, and that it lacked responsiveness to those who tried to report. Two Anglican chaplains were among the abusers at Dilworth School. An archbishop told the Inquiry that ordination changes the nature of the person — 'your being is changed' — and the Inquiry recorded that one minister's abuse had been described within the church as 'an affair'.
Presbyterian Church of Aotearoa New Zealand: a survivor's evidence that parents entrusted their daughters to the minister who abused them 'because he was a Minister', and an acknowledgement that Berhampore Home had no reporting system by which children's care could be monitored. This record is the thinnest of the seven and its scope note says so: it rests on one survivor's account and an oversight admission, not on a systemic doctrinal finding.
Two new categories: Anglican and Episcopal, and Methodist and Wesleyan. Around 18 Anglican records and 6 Methodist ones sat in the catch-all 'Other Christian' bucket while smaller groups — Lutheran with 5 records, Mennonite with 4 — each had their own. Existing records were moved across, reducing 'Other Christian' from 164 records to 143. This is the same granularity problem the July audit fixed between traditions, appearing inside Christianity.

Effect on published figures

Institutional-event records 131 to 137; sources 2,250 to 2,256; core conviction count unchanged at 1,030. Anyone grouping this dataset by Christian denomination should re-derive those counts: Anglican and Methodist records no longer appear under 'Other Christian'. All seven New Zealand records are inquiry findings, not criminal convictions, and none names an individual as a perpetrator except where that person's separate criminal record is cross-referenced. Where an institution disputed survivor evidence, the Inquiry's record of that dispute is carried in the record rather than resolved.

July 28, 2026 records

Waldorf record re-cut from the primary source and upgraded to a conviction; Plymouth Brethren inquiry finding added

An earlier version of the Waldorf record published today rested on secondary summaries and was filed as an institutional finding, with the teacher's convictions hedged out as unverifiable. Reading the underlying 200-page report in full changed that: it documents both convictions from the court's own files, and finds that anthroposophical doctrine — not merely the teacher's classroom position — strengthened his authority and supplied the frame in which the assaults were normalised. The record is now a conviction record. A Plymouth Brethren Christian Church finding from the New Zealand Royal Commission was added, and a cluster now links the Sri Lankan monastic ordination cases.

Germany, 1994: the Überlingen Waldorf record was rewritten from the IPP München report read in full, and re-filed as a criminal conviction rather than an institutional finding. Konrad Z. was convicted by the Amtsgericht Überlingen on 18 May 1994 of sexual abuse of persons entrusted to his care in conjunction with sexual abuse of children — 12 months suspended plus a DM 2,000 fine — over assaults on twin pupils in his own class. He had already been convicted in Switzerland on 20 August 1990 of indecency with a child at another Waldorf school, a conviction that became final, and had concealed those proceedings when hired. The school's board learned of it, accepted his account that it was a smear campaign after the Waldorf federation's legal adviser vouched for him as 'a completely upright person', and told neither the staff nor the parents of his class.
That record now rests on the doctrinal mechanism the report establishes, which is what brings it inside this archive's criterion. An ordinary school-abuse case fails the operative-means test: a teacher's authority is custodial, and a religious ethos in the surrounding institution is context, not instrument. The IPP's chapter on 'ideological risk conditions' found more than that — the doctrine that a pupil's placement with a teacher is a destiny relationship rather than chance 'considerably strengthens the authority of a class teacher', whose position it describes as often constructed as untouchable in the anthroposophical context; and a parent's evidence records that Z.'s sexual assaults were framed to the school community as fate and normalised, against a karmic premise under which no suffering is undeserved.
New Zealand, 2024: added the Royal Commission of Inquiry into Abuse in Care's findings on the Plymouth Brethren Christian Church, from 'Whanaketia', presented to Parliament on 24 July 2024. Thirty-two survivors told the Inquiry about abuse or neglect within the church's care, against five allegations the church acknowledges. The Inquiry found the church's belief that homosexuality is inconsistent with Biblical teaching, coupled with a belief that sexuality can be managed, led to attempts to 'correct' or 'manage' survivors' sexuality through conversion practices; recorded a member's evidence that a husband has complete authority over his wife 'in all matters including marital relations'; and found the church's assemblies self-regulating with no national policies and limited external accountability.
The Plymouth Brethren record carries the Inquiry's own record of the church's disputes rather than resolving them: the church does not accept the survivor evidence about sins being aired before other members, and while acknowledging five allegations does not necessarily accept they occurred within its care. It is an inquiry finding, not a criminal conviction, and names no individual as a perpetrator.
Added a cluster linking the five Sri Lankan cases in which a child was placed in a Buddhist temple to be ordained, or was resident there under a chief incumbent's care, and was abused by the monk holding that authority. Scholarly work on child ordination in Sri Lankan Theravada describes systemic barriers to disclosure around this pattern; the corpus records only the cases that reached an adjudicated outcome, which is a floor on the pattern rather than a measure of it.

Effect on published figures

Core records 1,029 to 1,030 and sources 2,246 to 2,250. The Waldorf record changed slug, class and evidence tier — anyone who retrieved it earlier today under the 2025 institutional-finding slug should re-fetch it. Also rejected this cycle and recorded as such: the Quaker material in the 2021 IICSA report, which is a finding about record-keeping and audit structure rather than about abuse enabled by spiritual authority, and names no perpetrator and no identified victim.

July 28, 2026 schema

Appellate verification is now stated on every conviction — including where it has not been done

A conviction that was later overturned does not meet this archive's criterion. Until now, a record with no appellate history meant two incompatible things — that the appeal record had been searched and nothing found, or that nobody had ever looked — and nothing in the data told them apart. Every record asserting a conviction now carries an explicit `appellateStatus`, and the proportion still unverified is published rather than left implicit. No record's evidence, sources, or proceedings changed.

Added `appellateStatus` to every record with `evidenceStatus: "convicted"`, with six values: `affirmed` (an appellate court considered it and the conviction stood), `modified` (conviction stood, sentence altered), `none-found` (searched, no appeal located — a checked state, not an unknown one), `disturbed` (conviction quashed, vacated or reversed), `pending` (appeal filed or within the appeal window), and `unchecked` (not yet verified). Records that do not assert a conviction carry null, because the question does not arise for a charge, a civil finding, or an official finding.
Published the current position in /data/COVERAGE.md: of 1,031 records recording a conviction, 272 (26%) have an established appellate position — 175 affirmed, 76 searched with no appeal found, 21 with the sentence modified while the conviction stood — and 739 (72%) are marked `unchecked`. Those are being worked by jurisdiction. An `unchecked` record is not a claim that no appeal exists; it is an open question, and the count of open questions is now stated so a reader can weigh it.
Recovered 76 records' worth of appellate research that the schema had been discarding. Researchers had already searched for appeals on these and written the result into `scopeNote` — 'No later appellate disposition was located', 'Appeal status unconfirmed' — because there was nowhere else to put it. That work was invisible to any query and is now a queryable state.
`disturbed` is never inferred automatically. Adverse language in a record's prose overwhelmingly describes something other than that record's own conviction falling: a co-defendant, a companion prosecution named for contrast, an extradition fight, or an appeal against sentence. Tested against the corpus, a keyword rule produced 38 matches that were mostly false positives — including the one Philadelphia conviction that stands, a conviction expressly upheld on appeal, a case whose murder convictions were reversed while the rape conviction it documents is live, and a case where an appeal court overturned a suspended sentence and jailed the defendant. A conviction that genuinely fell must be asserted by a person and is reviewed for removal or reclassification.
Wrongful-conviction records are excluded from derivation. The West Memphis Three record's own appellate history states that direct appeals affirmed the convictions in 1996 — true at the time, and followed by DNA exclusion and vacatur in 2011. Reading that as a standing conviction would invert what the record documents.
The field is exposed in the case index, the per-case API, the CSV exports and the OpenAPI schema, and is documented in `schema/CODEBOOK.md` §8. A build gate now requires an explicit value on every new conviction record; `unchecked` is permitted, so it never blocks a record being added — it only stops the backlog growing without being counted.
Corrected an overstatement in COVERAGE.md, which said a conviction's standing was established 'as documented at build time'. It now reads 'so far as the record establishes', which is what the data supports.

Effect on published figures

No records were added, removed, or reclassified. Anyone who has filtered this dataset on `evidenceStatus: "convicted"` and treated the result as convictions that stood should now also read `appellateStatus`: 26% of those records have a verified appellate position and 72% do not yet. The unverified majority is a gap in this project's verification, not a finding about the underlying cases.

July 28, 2026 records

Corpus enrichment: Zen, Afro-Brazilian, Hindu and evangelical records added; four duplicate records merged

A research cycle targeting under-covered traditions added five convictions and resolved four same-case-different-slug duplicates found during deduplication. Every added record rests on an adjudicated outcome in which the victim was inside the perpetrator’s religious authority. Candidates that failed that test are listed below as rejections, so the reasoning is public rather than invisible.

Germany, 2017: Hans Rudolf Döring (Dorin Genpo Osho), founding abbot of Bodaisan Shoboji — the first German branch temple of the Rinzai Zen Myōshinji line — and a vice-president of the World Fellowship of Buddhists, was convicted at the Landgericht Augsburg of serious sexual abuse of seven boys aged four to thirteen and sentenced to seven years and nine months. Two of the boys were the sons of a widow who had sought him out for grief counselling in his capacity as a Zen teacher; another was a refugee he had taken in through church asylum.
Scotland, 2025: Walter Masocha, self-styled archbishop and founder of Agape for All Nations Ministries International in Stirling, was convicted at the High Court in Livingston of attempted rape and of indecent and sexual assaults against two congregants, and sentenced to ten years in custody with a further four years of supervision. He told the women God had given them to him and that the contact was him casting out demons. A separate set of convictions against him from 2015 was quashed on appeal as a miscarriage of justice and forms no part of this record.
Brazil, Planaltina (Distrito Federal): a terreiro leader was sentenced to nine years, seven months and six days for the rape of a girl who was thirteen at the time and who attended the religious centre with her parents, themselves members. The sentencing court expressly found that he exploited the relationship of trust and the spiritual ascendancy — 'a ascendência espiritual' — that he held over her.
Brazil, Porto Alegre: a babalorixá was convicted of sexual violation by fraud against two women who consulted his religious centre, the acts carried out as what he presented to them as ritual procedure. Recorded as a first-instance conviction that has not become final, with appeal still available.
India, Sirsa district (Haryana), 2025: Raman Giri, priest of the Sherawali Mata temple at Ding village, was sentenced to 25 years of rigorous imprisonment under the POCSO Act, the Indian Penal Code and the IT Act for the sexual assault of a boy who had come to the temple to worship.
Four duplicate records were merged. Deduplication before writing found four same-case-different-slug pairs already in the corpus: Gurdarshan Singh, the Maryland granthi whose 1997 guilty plea was held twice; the Sorocaba pai de santo whose sentence was raised from 136 to 175 years on appeal; Tawanda Shumba, the Gweru shrine prophet, held twice under the same magistrate, date and sentence; and Mohammed Hanif Khan, the Bradford imam, held twice for the same Bradford Crown Court sentence of 13 years with a 3-year extended licence, one copy mis-dated to 2015. In each case the fuller record was kept and any distinct source from the other was folded into it. Duplicates of this shape pass the slug-uniqueness gate because both entries are individually valid; they are found only by scanning the corpus for matching names and places.

Effect on published figures

Tracker core records moved from 1,028 to 1,029 and sources from 2,241 to 2,246, after five additions and the removal of four duplicates. Two candidates were rejected on the criterion and are recorded as such: a Muzaffarnagar temple priest convicted of assaulting a boy he had lured away with the offer of a trip, where the priesthood was not the operative means of the coercion; and a Limpopo sangoma convicted of raping a ten-year-old who was a guest in his home as his wife’s niece, where the trust exploited was domestic rather than spiritual. Searches that returned no qualifying adjudicated outcome are also recorded, so they are not repeated blind: Diamond Way and Ole Nydahl, Kalu Rinpoche organisations, Ásatrú and Heathen kindreds in the Nordic countries, Wiccan covens in the UK and Australia, and Freemason and Odd Fellows lodges.

July 28, 2026 Classification

Classification audit: accusation violence separated from practitioner traditions

An internal bias audit of the classification layer found that the tradition vocabulary mixed three different relationships into single categories, and applied coarser labels and more delegitimising language to non-Western traditions than to Christianity. No case record's evidence, sources, or proceedings changed. What changed is how records are grouped and labelled.

Witchcraft- and sorcery-accusation killings are no longer filed under the tradition of the person who was accused. Ten records — where the victim was killed or persecuted for being accused of supernatural harm — moved to a new top-level category, 'Witchcraft- and sorcery-accusation violence'. Filed as before, a tradition filter counted these as offences by that tradition, when the tradition-identified party was the victim.
Satanic- and ritual-abuse panic prosecutions moved out of 'Western occult' into a new category, 'False ritual-abuse accusation and miscarriage of justice'. Eleven records — including the West Memphis Three, McMartin Preschool, the San Antonio Four, Orkney, Rochdale, and Outreau — document harm done to people wrongly accused, not by practitioners.
Every record now carries a `convictedParty` field, marking whether the convicted party held spiritual authority over the victim or not. Filtering the corpus for convictions previously returned records whose conviction ran opposite to the archive's thesis, with nothing in the data to distinguish them.
The tradition family label 'Indigenous spiritual authority or imitation' is now 'Indigenous spiritual authority'. Pairing a living tradition with 'imitation' in one label meant every fraudulent claim to Indigenous identity incremented that tradition's count.
'Claimed spiritual healing or mediumship' is now 'Spiritual healing, divination, and mediumship'. The hedge 'Claimed' appeared on the predominantly non-Western bucket and on no comparable Western one.
The tradition 'US curse-removal fraud' was retired and its records moved to 'Divination and fortune-telling services'. A tradition names what someone practised; the offence belongs in the record's type and proceedings, not in the identity axis.
The healing category was subdivided by practice — traditional practitioners, energy and bodywork, hypnosis and trance, mediumship and divination, plant-medicine ceremony, exorcism and spirit extraction — replacing a single undifferentiated bucket that held 68 records across 29 countries. Ten Western practitioners previously filed under 'New Age' moved into these shared practice buckets, so that comparable practice now sorts together regardless of region.
Buddhism, Hinduism, and Islam gained subdivisions comparable to those Christianity already had. Christianity was resolved to ten denominational buckets while Buddhism's records — spanning Theravāda, Vajrayāna, Zen, and Western convert sanghas across sixteen countries — shared one.
Type labels scoped to one tradition were rewritten to match actual use: 'Trafficking (juju/voodoo oath control)' is now 'Trafficking under spiritual coercion'; 'Curse-removal fraud' is now 'Fraud under spiritual pretext'; 'Initiation death' is now 'Death during ritual, initiation, or withheld care'. Each label named one tradition while the category was in use across ten or more.
The OpenAPI specification now derives its tradition and type enumerations from the live vocabulary instead of a hand-maintained copy, and declares the new `convictedParty` field. The `convictedParty` field was also added to the case index endpoint and to the CSV exports, alongside a derived `is_authority_conviction` column. Anyone counting convictions of spiritual authorities should use that column rather than `is_convicted`, which also counts convictions of accusing mobs and records of wrongful prosecution.

Effect on published figures

Record counts per tradition changed for the categories listed above. The corpus totals — 1,283 records, 1,028 core, 2,241 sources — are unchanged, as is every record's evidence, sourcing, and proceedings history. Anyone who cited per-tradition counts before this date should re-derive them from the current dataset or the versioned API.