No Way Out: How Russia’s Courts Convict Almost Everyone Charged Under Its Speech, Protest and Terrorism Laws
1. Publication. This report is published by Trace Origin LLC (“Trace Origin”), an independent research organisation, on 17 October 2026 (version 1.0) and revised on 27 October 2026 (version 1.1). The analysis, characterisations and opinions in it are Trace Origin’s alone. OVD-Info provided data used in the report, with permission, and is not responsible for the analysis. Any later correction is dated and recorded (section 7).
2. Sources. The report reproduces and summarises official public records: case pages published by courts of the Russian Federation on their official websites (sudrf.ru), as retrieved on 15–17 October 2026, with the retrieval time and a SHA-256 checksum preserved for each page. It also draws on OVD-Info’s records and on public sanctions data compiled by OpenSanctions, and on published national statistics (Supreme Court and Judicial Department data as reported by RBC, The Moscow Times and SOVA Center) and on press reports (Fontanka). Every statement of fact about a court, a case or a judge is attributable to one of these sources, and is identified as such. Trace Origin does not warrant the accuracy of the underlying records; courts may amend or withdraw them.
3. Statements of fact and of opinion. Figures, dates, case outcomes and the names of presiding judges are reported as they appear in the cited records. The title, headings and characterisations of patterns (for example, that the way out of a prosecution is “closed”) are the author’s opinion and analysis, based on the facts disclosed in this report, which the reader can check against the sources. References to international law are the author’s analysis and not the finding of any court or treaty body.
4. No allegation against any individual. Nothing in this report states, or is intended to imply, that any named person has acted unlawfully, corruptly, in bad faith or in breach of any professional duty. Judges are named only because the courts’ own public records name them as presiding over a case, in their official capacity. Statistics about outcomes, custody or the assignment of cases describe patterns across many records and do not state that any individual trial was unfair or any individual decision wrong. Where a panel of judges sat, the record names only the presiding judge, and the decision is the panel’s.
5. Sanctions. A statement that a judge’s name matches a sanctions listing means only that the name, and where stated the court, correspond to an entry in public sanctions data. Unless the report states that a match has been confirmed against the Official Journal of the European Union or another official publication, it does not state or imply that the person is subject to sanctions. Sanctions designations are preventive administrative measures and are not findings of guilt.
6. Personal data. Personal data are processed for purposes of journalism and research in the public interest, limited to what those purposes require. The report names no defendant, lawyer, victim or witness. Judges and prosecutors are named only as public officials, in relation to official acts, from the courts’ official publications. This published version omits case numbers, links and other details that would lead a reader to a defendant; those are shared only with OVD-Info. Requests concerning personal data, including access, rectification and objection, may be sent to ebrattin@traceoriginresearch.com.
7. Right of reply and corrections. Any person named in this report may submit a response or a request for correction to ebrattin@traceoriginresearch.com. Trace Origin will review each request promptly, correct any verified error, record the correction with its date, and, where a response is received, publish or summarise it alongside any published version.
8. No advice; limitation of liability. This report is not legal advice and does not create any professional relationship with any reader. It is provided “as is”, without warranty of any kind as to completeness or accuracy. To the extent permitted by law, Trace Origin accepts no liability for any loss arising from reliance on it.
9. Third-party data. OVD-Info data are used with permission and are not redistributed. Court directory data: Dataout / sudrfparser, licensed CC BY-SA 4.0. Sanctions data: OpenSanctions, used under its licence terms.
No Way Out
How Russia’s courts convict almost everyone charged under its speech, protest and terrorism laws
Cases that reach these courts almost never end in acquittal, and rarely in dismissal, even where the charge does not rule it out. The courts' own records show where these cases are tried, and name the judges who try them.

This report reads the courts' public case pages for 3,487 first-instance cases, almost all registered from 2022 to 2026, under the articles listed in Annex A, and checks them against OVD-Info's records. The cases are chosen by article, not by motive: alongside anti-war speech and protest they include calls to violence, armed attacks and treason, and the case pages do not allow them to be separated. The report covers first-instance courts only. It follows a defendant through the system: how the case ends, how long the defendant is held, which court hears it, how it is handled alongside others, who presides, whether those judges appear on sanctions lists, and what OVD-Info's records are missing. The court pages matter more now: according to Fontanka (3 October 2026), the Judicial Department's summary table of 2025 criminal statistics leaves out the terrorism and extremism articles, treason and 280.3.
(1 of 2,292 convicted or acquitted;
district and regional courts 0.6%; all courts 2022: 0.3%)
does not rule it out (anti-war articles,
lesser parts, only charge)
article 205.2, a terrorism offence
committed by speech, as the only charge
The argument in seven findings
- 1Almost no one is acquitted, and few cases are dismissed. Of defendants convicted or acquitted, military courts acquitted 0.04% (1 person), district and regional courts 0.6% (3); all Russian courts acquitted about 0.3% in 2022. Where the charge does not rule out dismissal, it is rarely recorded: 4 of 136 defendants charged only under the lesser parts of the anti-war articles. The commonest outcome other than conviction is a compulsory psychiatric measure: 93 defendants in military courts, 69 of them under 205.2.
- 2Custody is extended hundreds of times while cases wait. The case pages log 901 decisions extending custody. Of eight people OVD-Info listed as imprisoned in August 2026 whose case showed no first-instance decision in mid-October, six still had none recorded; one case was refiled and decided, and one moved to another court and must be traced there.
- 3Expression charged as terrorism is tried in military courts. 1,621 of the 2,846 military-court cases include article 205.2 (public calls to, justification of, or propaganda of terrorism); on the pages read, it is the only charge in 963 registrations. Its online form, part 2, carries more than five years, which removes most routes out (Finding 1).
- 4Cases are handled as a group. One district court received the first materials of ten cases that OVD-Info records separately in two runs of nearby case numbers, on three days, and listed them for trial in the same order. A search of every case page read finds no other court where such a run is statistically significant.
- 5The records name who presides, and some judges' case pages record one uniform sequence. Every case page read names the presiding judge. The cases are spread across many judges: within each type of court, the 10% with the most cases presided over about a quarter of them (28% in military courts, 24% in district courts). For 5 military-court judges, most case pages on given articles record one short, uniform procedure, with sentence 24 to 38 days (median) after the case reached the judge; their colleagues' records vary far more.
- 6Some of those judges match a sanctions listing; most of the busiest do not. 9 judges (one a weak match) match an EU listing that describes a judge. 32 more, 29 of them at the Southern District Military Court, share a full name with a person on Ukraine's sanctions register. All are unconfirmed. Of the 21 judges recorded as presiding over cases with 25 or more convicted defendants, 3 match one or the other.
- 7What the court records show OVD-Info is missing. For 207.3, 280.3 and 284.1, OVD-Info's records appear nearly complete where courts publish names. Under 280.4, 15 of 22 defendants were not matched (about 11 allowing for missed matches), and many recent cases are missing.
Then: what these findings do not establish; Annex A, method; Annex B, tables for follow-up; questions and answers.
What was examined, and how
The source is the court's public case page (карточка дела) on its website (sudrf.ru): the judge, every procedural event, and the result for each defendant. Each page was fetched once, stored with a SHA-256 checksum and the time of retrieval, and read offline between 15 and 17 October 2026. Pages were read in full for 3,316 of 3,487 cases. Courts withhold some pages (closed hearings, state secrets); outcomes in those cases are unknown and may differ from the rest.
| Courts | How they were searched | Cases | Case page read in full |
|---|---|---|---|
| District military courts | All of them, nationally, under the articles in Annex A | 2,846 | 2,725 |
| District and regional courts | Every court in the 23 census regions (Annex A) under five articles, plus searches for OVD-Info cases in other regions | 641 | 591 |
The two groups of courts were searched differently, so their totals cannot be compared or simply added, and the report gives figures for each. Finding 2 relies on 172 OVD-Info records that were matched to a court case in four regions. Finding 4 started from those records and was then tested against every case page read. Case counts include cases whose page was withheld; outcomes, events and durations cover the pages read. Finding 7 relies on the census of 23 regions.
The report names no defendant, and this published version gives no case number or link that leads to one.
Cases that reach these courts almost never end in acquittal, and rarely in dismissal, even where the charge does not rule it out
Of 2,849 decided defendants, 2,822 were convicted, 23 had the case dismissed and 4 were acquitted. The acquittals were 3 at a regional court and 1 at a district military court. In the military courts: 1 acquittal among 2,302 decided defendants and 10 dismissals; in the district and regional courts, 3 acquittals among 547 and 13 dismissals. As a share of defendants convicted or acquitted, the military courts acquitted 0.04% and the district and regional courts 0.6%. Across all Russian criminal cases in 2022, first-instance courts acquitted 2,062 people against about 614,700 convicted, about 0.3% on the same arithmetic (Supreme Court data, as reported by RBC and The Moscow Times). The national figure also covers magistrates' courts and private prosecutions, which this report does not, so it is a reference point, not a like-for-like comparison: the military courts acquit less often than it; the district and regional courts, on small numbers, somewhat more often.
The law leaves other ways out. A case can end without conviction on non-exonerating grounds. A court fine in lieu of prosecution and general active repentance are open to first offenders charged with offences of small or medium gravity (a maximum of five years or less). For several graver articles, notes to the article itself allow release from liability, for example for voluntarily reporting or ceasing participation (205, 205.1, 205.4, 205.5, 208, 275, 282.1, 282.2). Limitation applies at any gravity. Reconciliation needs a victim; most offences of expression have none, though insulting veterans (354.1 part 3) and offending believers (148) can.
Where the charge does not rule it out, dismissal is rare. The charge only opens the door: each route has further conditions (a first offence, reparation of harm, the defendant's consent, a request by a party), and the case page does not show whether a defendant met them or whether anyone asked. What is measured is how often dismissal was recorded where the charge did not rule it out. Counting defendants whose only charge is the article part in question: under the lesser parts of the anti-war articles (207.3 part 1, 280.3 part 1, 284.1 part 2), which carry five years or less, 4 of 136 cases were dismissed (2.9%), all in district and regional courts. Under 205.2 part 1 in the military courts, 0 of 57. Among all defendants charged, 280 part 2 in the military courts: 1 of 212; 205.5 part 2, an article with a release note: 5 of 558. Under 354.1, parts carrying up to three years ended in dismissal for 7 of 33, parts carrying up to five years for 0 of 100; both are eligible, so the difference is practice, possibly reconciliation with a veteran as victim.
No reliable national comparison yet. National figures by article exist for 2023 (SOVA Center, from Judicial Department data): dismissed as a share of convicted or dismissed, 205.2 1.1% (1.4%); 282.2 2.7% (3.3%); 280 2.9% (4.6%); 282 5.6% (7.7%); 354.1 12.0% (16.1%) (main-charge convictions only in brackets). The shares here are lower for all five (205.2 4 of 1,332 (0.3%); 280 1 of 243 (0.4%); 282.2 0 of 118 (0.0%); 282 0 of 22 (0.0%); 354.1 7 of 163 (4.3%)), but the comparison does not yet measure like with like. National tables count each article charged, so a defendant convicted on one article and released on another counts as a dismissal; here a defendant counts as dismissed only if the whole case ends. For 205.2, heard at first instance only by the district military courts, this report covers essentially every case in the country, so a difference can come only from the years covered (2022–2026 against 2023) and from how outcomes are counted. A like-for-like test (2023 decisions, one charge, defendants present in court) is in preparation; until then this report makes no claim that these courts dismiss less often than others.
The commonest outcome other than conviction is psychiatric. Compulsory medical measures (Criminal Code arts. 21 and 81; CPC chapter 51) are ordered by a court decision, not a verdict. A person found not responsible at the time of the act is released from criminal liability; a person who fell ill afterwards is released from punishment, and can be prosecuted on recovery. Treatment, in hospital or as an outpatient, has no fixed end and is reviewed by a court. It is not a dismissal and not a way out in the ordinary sense. The case page records that measures were applied, not on which ground. This report does not count it as a decision. On the case pages read it was ordered for 93 defendants in the military courts, 69 of them under 205.2 (4.9% of 205.2 defendants with a final outcome, close to the national share of 5.7%), and 18 in the district and regional courts. Nationally in 2023, 23 people charged under 205.2 were sent to compulsory treatment, against 4 dismissals and 375 convictions.
Acquittal is as rare here as in Russian criminal justice generally, and rarer still in the military courts. Dismissal, too, is rarely recorded where the charge does not rule it out; whether that is rarer than in other Russian cases is not yet measured (above). Dismissal under the lesser anti-war charges and release under the notes to the terrorism articles are rare. Dismissal at trial usually follows a motion by the prosecution or a party, and cases the investigator ends never reach court, so these rates describe the whole chain, not the judge alone. They fit what research on Russian criminal justice calls accusatory bias: acquittals and dismissals count against investigators' and prosecutors' performance indicators, weak cases are filtered out before trial, and judges avoid reversal on appeal (the Institute for the Rule of Law at the European University at St Petersburg; Volkov, Paneyakh and Titaev; Solomon; Hendley). On that reading these cases sit at the far end of a general pattern; what is specific to them is the charge structure described in Finding 3. Charging online speech under 205.2 part 2, which carries more than five years, also removes the court fine and general active repentance from the start. A non-exonerating dismissal is not a finding of innocence: it ends the prosecution, usually with the person's consent and acceptance of the facts, and is a way out only in that sense. Compulsory medical measures, the commonest outcome other than conviction, end the case without a verdict but impose treatment of no fixed length.
Gravity classes rest on our reading of the maximum sentence for each part, which must follow the law at the time of the offence: the penalties for 207.3 and 280.3 changed in 2022–2023, and the version used here should be confirmed by counsel, as should the scope of each article's release note. National figures are for 2023 only; this report covers 2022–2026. Defendants tried in absentia (CPC art. 247(5)), common among emigrants charged under 207.3, 280.3 and 284.1, cannot meet some conditions for dismissal and are not separated here; the case page often shows it, and the next run will. Per-part counts can include a defendant charged under two parts twice, and the outcome counts here are defendant rows on the pages read, so they differ slightly from the court-level totals above, which merge refiled cases. Numbers where the charge permits dismissal are small, and the individual conditions for it are not observed. Nothing here shows that any trial was unfair: the presumption of innocence (ICCPR art. 14(2)) places the burden of proof on the prosecution and requires that the accused be treated accordingly, and a conviction rate alone is not evidence of its breach. Cases still pending and withheld pages are not counted.
The case pages record 901 decisions extending custody, an undercount; six people OVD-Info listed as imprisoned have no decision recorded
The case pages log 901 decisions extending custody, 848 in the military courts. These are mostly extensions during trial: extensions during the investigation are filed as separate court materials and largely do not appear on the case page, so the count understates custody. The page does not show how many defendants were held, so the count has no per-person denominator. In 414 cases the sentence came on the only hearing day the page records (a hearing and a sentence on the same day count as one; Finding 5 looks at who decides these cases and how fast).
Among the OVD-Info records matched to a court case, eight people listed as imprisoned have no first-instance decision on the case page as of 15–16 October 2026. Six show no decision, transfer or return; their prosecutions began between 2022 and 2025, by OVD-Info's dates; the court record cannot show whether they were still held on that date. For the other two (the last entry is a return of the case to the prosecutor or a transfer to another court), the graph follows the case identifier across courts: the returned case was refiled and decided, so that person is not waiting; the transferred case has not yet been traced to its new court.
Article 9(3) of the ICCPR provides that detention pending trial shall not be the general rule. That is a question about the share of the accused who are held, which the case page cannot answer; what it does show is custody renewed by the court hundreds of times in these cases, and people OVD-Info lists as imprisoned whose prosecutions began one to four years ago and who still have no first-instance decision. Single-hearing sentences are not long trials with missing entries: Finding 5 shows they come within weeks of the case reaching the judge. Since 2020 the shortened guilty-plea procedure (chapter 40 of the Code of Criminal Procedure) is limited to offences of small and medium gravity, so for most charges here the likely explanations are a cooperation agreement with the prosecution (chapter 40.1) or a trial in the ordinary procedure with an admission and the evidence read in one sitting.
A count of extension decisions is not a count of people or of days in custody. The time shown for the people waiting runs from the start of the prosecution, not from the case reaching court, and whether a person is imprisoned comes from OVD-Info's export of August 2026, not from the court record. The eight come from matched records in four regions, not the census.
Article 205.2, which the Criminal Code places among terrorism offences although it punishes expression, is in 1,621 military-court cases, and the only charge in 963 of those read
District military courts sit at the level of regional courts and hear terrorism cases at first instance, for civilians as well as servicemen, each for a district of several regions. They are federal courts within the ordinary judiciary, not courts-martial. Of their 2,846 cases examined, 1,621 include article 205.2: public calls to terrorism, public justification of terrorism, or terrorist propaganda. All three are offences of expression. On the 2,739 military-court registrations whose case page was read (refiled cases appear more than once, so this differs from the 2,725 cases above), 1,607 include 205.2, and in 963 it is the only article charged. 1,494 cases include other terrorism offences, treason or espionage; a case can include both.
The courts differ. At the Central District Military Court (Yekaterinburg), 565 of 856 cases include 205.2. At the Southern District Military Court (Rostov-on-Don), 443 of 588 include other terrorism offences, treason or espionage and 184 include 205.2. The Southern court hears cases from the North Caucasus and from occupied Crimea, and has tried captured Ukrainian servicemen, so its caseload is of a different kind; the case page does not say which defendants these are.
Russian law assigns these articles to the district military courts (Code of Criminal Procedure, art. 31): the jurisdiction is authorised by domestic law, which says nothing about whether any individual proceeding complied with it. Under the ICCPR they are contestable: article 14(1) requires a competent, independent and impartial tribunal, and the UN Human Rights Committee's General Comment No. 32 (para. 22) treats the trial of civilians by military courts as exceptional, limited to cases where the state shows it is necessary and justified by objective and serious reasons and where the regular civilian courts are unable to undertake the trials. Incitement to terrorist violence can be lawfully punished (ICCPR art. 19(3)); the question in each case is what was said. The choice of 205.2 over speech articles that carry lower penalties (280, 280.4, 207.3) is the prosecutor's; how often it is made cannot be measured from the case pages. Since 2009 the judges of military courts are no longer in military service (Federal Constitutional Law No. 3-FKZ of 29 June 2009), though many are former officers; whether that answers the concern in General Comment No. 32 is a question for counsel. Part 2 of 205.2, the online form, carries more than five years, which removes the court fine and general active repentance as ways out; the ways out that remain are seldom used (Finding 1).
The case page does not show what was said, and many 205.2 judgments are unpublished. The military courts' caseload includes calls to violence and jihadist propaganda as well as anti-war posts. The page does not record whether a defendant is a civilian, a resident of occupied territory or a prisoner of war. Being tried in another region is the normal result of a military court's multi-region district, not a finding.
One court received ten cases that OVD-Info records separately on three days, in two runs of nearby case numbers, and listed them for trial in the same order
OVD-Info records ten cases at a district court in Krasnodar Krai, all under article 282.2 (an "extremist" organisation), each as a separate prosecution. The court's pages record the seven earlier cases as received on two consecutive days at the end of 2023, under case identifiers within a run of nine, and the three later cases on a single day in 2024, within a run of twelve. On the court's 2025 trial list they appear as two runs of consecutive dockets (eight dockets for the seven 2023 cases, three for the 2024 cases), in the order received, divided among three judges.
No other significant run was detected. The pattern was first found by hand, so the test below is exploratory, not confirmation. To test whether it is unusual, each of the 3,333 court registrations whose case page was read (a refiled case can have more than one) was grouped with others whose case identifiers lie within 20 numbers at the same court and year, under a shared article, and whose case pages record receipt within 45 days of each other. Each group was scored against tracked cases falling at random in that court's sequence, correcting for the number of groups tested. Of 466 groups of two or more, two are significant: the 2023 and 2024 runs at this court. Eleven identifier numbers inside the two runs belong to no case in our data: other materials, possibly concerning other people. Near misses elsewhere (corrected q between 0.05 and 0.25) are listed for OVD-Info but not reported here. Because the run was noticed first and then tested on the same records, the test shows how unusual it is in these data; it is not an independent discovery. The test also assumes cases fall at random in a court's sequence, while filings under one article may cluster for ordinary reasons.
Consecutive receipt of first materials, usually requests to detain or search, is consistent with one coordinated operation, or with one criminal case split into several (Code of Criminal Procedure, art. 154); the case files would show which. For OVD-Info the practical point is one docket inside the sequence with no OVD-Info record: possibly someone not yet documented.
Handling cases together says nothing about how carefully each was tried. That OVD-Info records the cases separately means only that no link was recorded, not that the cases are unrelated. The search covers only cases under the tracked articles whose case page was read, so batches of other cases are not visible to it.
Every case page read names the presiding judge; most judges share the work, and for some the record shows one short, uniform procedure
The case pages name 660 presiding judges (counted per court). Within each type of court, the 10% of judges with the most cases presided over about a quarter of them: 28% in military courts (20 of 208 judges), 25% in regional courts and 24% in district courts. At the six busiest military courts, the three judges with the most cases presided over 15% to 34% of each court's cases examined.
| Presiding judge | Court | Cases | Convicted / acquitted | Custody extensions |
|---|---|---|---|---|
| Shelestov Andrey Nikolayevich | Central District Military Court | 70 | 48 / 0 | 21 |
| Kakadeyev Anton Igorevich | 2nd Western District Military Court | 48 | 46 / 0 | 10 |
| Kolesnikov Aleksey Valeryevich | Central District Military Court | 44 | 44 / 0 | 26 |
| Solomko Igor Ivanovich | Central District Military Court | 63 | 42 / 0 | 23 |
| Massin Yuri Aleksandrovich | 2nd Western District Military Court | 38 | 42 / 0 | 7 |
| Slepukhin Andrey Vasilyevich | Central District Military Court | 43 | 41 / 0 | 0 |
| Nikitchuk Igor Vladimirovich | 2nd Western District Military Court | 29 | 37 / 0 | 10 |
| Shishov Oleg Aleksandrovich | 2nd Western District Military Court | 37 | 34 / 0 | 22 |
| Pluzhnikov Andrey Viktorovich | 2nd Western District Military Court | 35 | 34 / 0 | 7 |
| Belkin Igor Viktorovich | Central District Military Court | 39 | 31 / 0 | 14 |
As recorded on the courts' case pages. The ten judges recorded as presiding over the most convicted defendants (three share 31 at tenth place).
Judges whose case pages record one uniform sequence. Each case page records the steps of the case in order. Reduced to that sequence (registered, sent to the judge, hearings held or postponed, custody extended, sentence), each judge's cases on one group of articles were compared with those of the other judges at the same court on the same articles, by permutation test, correcting for the 69 judge-and-article groups with ten or more decided cases. Comparing within a court removes differences in how each court records its events. In 8 groups a far larger share of cases follows a single pattern than among colleagues' cases. One judge whose 43 case pages record no custody decision of any kind is left out, because his pattern is most likely record-keeping; 6 groups, covering 5 judges, remain:
| Presiding judge | District military court | Articles | Decided cases | Usual pattern | Share on that pattern | Colleagues: share on their most common pattern | Median days, judge to sentence |
|---|---|---|---|---|---|---|---|
| Judge A | 2nd Western | 205.2 | 14 | one hearing, then sentence | 86% | 32% | 27 |
| Judge B | Central | 205.2 | 11 | sentence with no earlier hearing recorded | 73% | 17% | 35 |
| Judge C | 2nd Western | 205.2 | 14 | sentence with no earlier hearing recorded | 71% | 35% | 24 |
| Judge D | Southern | terrorism, treason, espionage | 11 | one postponed hearing, then sentence | 64% | 21% | 38 |
| Judge E | Central | 205.2 | 13 | sentence with no earlier hearing recorded | 62% | 17% | 29 |
| Judge A | 2nd Western | terrorism, treason, espionage | 25 | one hearing, then sentence | 56% | 13% | 31 |
Every case on these patterns was sentenced within 90 days of reaching the judge, at a median of 24 to 38 days. Across all courts, the 368 cases sentenced with no earlier hearing recorded took a median of 24 days; the 192 cases with three or more hearings held and no other step recorded took 88; more hearings take more days, so this contrast describes the patterns rather than measuring haste. (The 368 are cases whose page records no hearing before the sentence; the 414 in Finding 2 count a single hearing day, including a hearing and a sentence on the same day.) The time colleagues at the same courts took on the same articles was not compared.
Part of this uniformity may be record-keeping. Each judge's own staff keeps the case page, and the step in which the judge schedules the trial, which must precede it, is often missing. The judge left out above is the clearest case. The dates of receipt and sentence are recorded either way, and the speed rests on them. The judges are identified by letter here and named in Annex B5, because a uniform record describes how the case page was kept, not how the judge conducted the trial.
Source: the courts' case pages; analysis of 17 October 2026. Statistical significance: q < 0.01 for every group shown, after correction.
The case page links a named presiding judge to a case and its result. That link does not by itself establish individual responsibility, which under the ICCPR rests with the state and, for individuals, only under specific regimes. How cases are assigned to judges was not examined. A uniform record may reflect how a court's staff enter events as much as how hearings were run, and an entry missing from the page is not a hearing that did not happen. A trial completed within weeks, at the first sitting the record shows, is consistent with a cooperation agreement (CPC chapter 40.1; since 2020 the chapter 40 guilty-plea procedure is closed to grave offences) or an admission with the evidence read into the record in one day; the speed of the trial stage says nothing about the defence's preparation time, which runs from the charge and access to the case file (CPC art. 217).
Where a panel of three sat, the page names only the presiding judge, and a panel verdict is a collective decision: counts per judge measure how often a judge presided, not individual responsibility for the result. Whether cases are steered to particular judges needs each court's ordinary caseload, which has not been collected. A short record is not evidence that no hearing took place. Whether these were cooperation agreements, guilty pleas or contested trials is stated in the judgments, which have not been read.
9 judges match an EU listing that describes a judge; 32 more share a full name with a person on Ukraine's sanctions register
Every judge was checked against OpenSanctions, which consolidates EU, UK, US, Ukrainian and other lists. Two kinds of match are reported, and none is confirmed.
EU listings that describe a judge. Same surname and initials, the same first name and patronymic where the page gives them, and a listing that describes the person as a judge: 9 judges, including one weak match on a common name, at 10 courts (one judge, recorded once by full name and once by initials, is taken to have sat at two). Of these 10 judge-court pairs, 8 listings name the same court and 2 name a different one. Together they were recorded as presiding over 97 of 3,487 cases.
Ukraine's sanctions register. 32 further judges share surname, first name and patronymic, allowing for Ukrainian spelling, with a person on the State Register of Sanctions kept by Ukraine's National Security and Defence Council. The register's entries give a decree and a date of birth but not an occupation, so the listing alone cannot show that the person is the judge. Entries that also appear on another list are excluded, because in the two such cases checked the other list described a different person (an EU listing of a Donbas separatist, an EU Libya listing); one judge whose entry is also on the Anti-Corruption Foundation's list is set aside for review. 29 of the 32 sit at the Southern District Military Court. Together they were recorded as presiding over 517 cases, with 371 convicted defendants.
21 judges were recorded as presiding over cases with 25 or more convicted defendants, 709 in all. 2 match an EU listing and 1 a full name on Ukraine's register; the other 18 match neither.
EU restrictive measures and Ukrainian sanctions are preventive measures, not findings of guilt. The EU listings matched here were made under three regimes (Annex B). If the matches on Ukraine's register are confirmed, Ukraine has listed 29 of the 40 judges recorded at the Southern District Military Court in these data. The likely reason, to be checked against the decrees, is that the Southern court tries residents of occupied Crimea and captured Ukrainian servicemen.
The EU has tended to list judges from prominent trials rather than by caseload; these data are consistent with that, but do not test it.
Every match is a lead. EU listings must be confirmed against the Official Journal; matches on Ukraine's register, against the date of birth and the decree. A shared full name is not proof of identity. A judge with no match may be listed under a spelling the matcher missed; 734 other name candidates, 348 of them on a sanctions list, did not meet either standard and are unreviewed. Lists outside OpenSanctions were not checked.
OVD-Info's records appear nearly complete for 207.3, 280.3 and 284.1, but not for 280.4 or for recent cases
Every district and regional court in the 23 census regions was searched under four articles with no ordinary-crime use, 207.3 (army "fakes"), 280.3 (discrediting the army), 280.4 (calls against state security) and 284.1 ("undesirable" organisations), and under 354.1 ("rehabilitation of Nazism"), which has one. Counting only searches run to the end, merging cases registered twice, and leaving out 354.1 part 2 (below), the courts list 329 defendants; for 100 the court withholds the name. OVD-Info has a record for 171 (52%; 95% interval 47–57%, somewhat too narrow because co-defendants in one case are counted as independent). Each defendant was checked against OVD-Info by docket or, failing that, by surname, initial, region, article and date.
The anti-war articles. OVD-Info has 47%–81% of the defendants the courts list under the four largest anti-war article parts (207.3 part 1, 36 defendants; 207.3 part 2, 49 defendants; 280.3 part 1, 70 defendants; 284.1 part 2, 59 defendants). Where courts publish names, OVD-Info has 71%–88% of them before any correction. The name check misses some true matches (it found 63% in a test on 43 people with a known docket), so these are lower bounds. Scaling them up by that rate gives 100% or more for several parts, which shows the correction is too crude for a point estimate: the gap under these articles is probably small, but it cannot be measured more precisely here.
280.4. The courts list 22 defendants under 280.4 part 2; OVD-Info has 7 on record. The other 15 come from 17 registrations, two pairs of which are one case each, registered twice. Allowing for matches the name check misses, about 11 of them are likely absent from OVD-Info. Annex B3 gives totals; the case details are shared with OVD-Info only.
Recent cases. Where courts publish names, OVD-Info has 53% of defendants registered in 2022, 77% in 2023, 86% in 2024, 84% in 2025, 36% in 2026, before any correction. The 2026 figure partly reflects the date of OVD-Info's export (August 2026): recent cases have had less time to be documented. Coverage is also lower for 2022, the first months of the new articles.
354.1 part 2 is mostly outside OVD-Info's scope. It is left out of the figures above: 38 defendants after merging cases registered twice. For 18 of those not found in OVD-Info a judgment was published and read by hand: 11 concern Holocaust denial, antisemitism or admiration of Hitler; 2 concern claims about Soviet history and should be checked; 5 could not be classed. The rest are unread.
For OVD-Info this is the actionable part: a list of court cases to verify, and an outside check on its documentation from the state's own records.
Where the court withholds a name, coverage can be checked only by docket, so the gap there is unknown. A person missing here may be in OVD-Info under another article or spelling. The correction for missed matches exceeds 100% for several article parts before it is capped, so it is a rough guide, not a measurement. The census covers district and regional courts only. Searches not run to the end (233 registrations) are left out because they over-represent courts where OVD-Info already has cases.
What these findings do not establish
- Not motive. The population is defined by article and includes genuine terrorism. It is not a count of politically motivated prosecutions.
- Not a before-and-after. Near-certain conviction is long-standing. Several articles predate 2022: 205.2 (2006), 282.2 (2002), 280.1 (2013), 354.1 (2014) and 284.1 (2015); 207.3, 280.3 and 280.4 were created in 2022. The military courts' terrorism jurisdiction also predates 2022, and the district military courts were reorganised and renamed in 2019. This report has no pre-2022 comparison.
- Not the whole country. Military courts were searched nationally; district and regional courts in 23 census regions, chosen for the study rather than at random. Garrison military courts were not searched. Totals for the two groups cannot be compared or added.
- Not the European Convention. Russia was bound by the European Convention on Human Rights until 16 September 2022, and the European Court of Human Rights still decides applications about facts before that date, which covers the first months of prosecutions under 207.3 and 280.3. This report analyses the cases under the ICCPR only.
- Not international humanitarian law. Residents of occupied Crimea and captured Ukrainian servicemen tried at the Southern District Military Court are protected by the Geneva Conventions (the Fourth for civilians in occupied territory, the Third for prisoners of war), which this report does not apply. The case page does not identify them.
- Not the fairness of any trial. The findings describe official records, not what happened in court.
- Not individual responsibility. The page names a presiding judge. Panels decide collectively; responsibility under the ICCPR rests with the state.
- Not an identification on a sanctions list. Every match is a lead pending confirmation: EU listings against the Official Journal, Ukrainian ones against the date of birth and decree.
- Not only 2022 onwards. A small number of cases registered before 2022 (at least 26) are included where the searches returned them.
- Not the appeal. Only first-instance outcomes were read. Appeals and cassation, which can change a verdict or a sentence, are not covered, so “no way out” describes the trial, not the whole process.
- Not every case. Courts withhold some pages, often in treason and closed cases; those outcomes may differ.
- Not the prosecutors. Courts withhold the prosecutor's name on about 95% of these case pages ("information hidden", or the office only), so whether particular prosecutors and judges are paired could not be tested.
- Not independent of OVD-Info. OVD-Info's records (export of August 2026) supply the matched cases, the information that a person is imprisoned, and the start dates; the court pages are the independent check.
How the data were collected and checked
This annex is written for technical reviewers.
Collection
- Search pages are behind captchas, typed by a person; nothing was solved automatically. Case pages need none.
- Articles covered: 207.2, 207.3, 280, 280.1, 280.3, 280.4, 282–282.4, 284.1, 284.2, 330.1, 354.1, 148, 128.1, 212.1, 329, 205–205.5, 275, 275.1, 276, 278, 279.
- Every tab of each case page is parsed: the case (UID, judge, single judge or panel), progress (each event with date, time, result and basis), persons (result per defendant; names are not kept from case pages), parties (role only; prosecutors named as officials). The coverage check in Finding 7 uses defendants' names from the courts' search results and OVD-Info's records; they are kept on Trace Origin's computer only and never published.
- Census regions (23): Sverdlovsk, Bashkortostan, Moscow Oblast, Orenburg, Krasnodar, Arkhangelsk, Omsk, Nizhny Novgorod, Tatarstan, Buryatia, Novosibirsk, Altai Krai, Bryansk, Adygea, Murmansk, Astrakhan, Perm, Belgorod, Altai Republic, Chelyabinsk, Novgorod, Nenets, Oryol. Check of the census: all 42 OVD-Info records in the first four regions that had already been matched to a district or regional court under the census articles were found by the search.
- Coverage: docket match where OVD-Info has a docket; otherwise exactly one OVD-Info record in the region with the same surname and initial, a shared article and a plausible date. The name check finds 63% of known matches (tested on 43 people with a known docket).
- Sanctions: OpenSanctions, 17 October 2026; patronymics in Russian, Ukrainian and Estonian spellings; both initials and, where given, first names must agree; the listing must describe a judge. Ukrainian register entries carry no occupation, so they are reported separately when surname, first name and the full patronymic agree (allowing for Ukrainian spelling) and, as a conservative rule, the entry is on no other list.
Graph analyses
- The waiting count in Finding 2 comes from one RDF graph joining Trace Origin's OVD-Info persecution graph to the court records: registrations, cases keyed by case identifier (УИД), courts, regions and presiding judges. The waiting test is a closed-world query (SPARQL), not an OWL inference: OWL's open-world semantics cannot conclude that a result is absent. The batch search and the procedural templates run directly on the case-page data.
- People waiting (Finding 2): an imprisoned person counts as waiting only if no registration of the same case identifier, at any court, has a final result. This follows a case through transfers and refiling. Waiting is defined relative to the pages read: a decision on a withheld page or at an untraced court would not be seen.
- Batch search (Finding 4): groups of cases at the same court and year whose identifiers lie within 20 numbers, which share an article and whose case pages record receipt within 45 days of each other; each group scored against random placement of the tracked cases in that court's sequence (scan statistic), with Benjamini–Hochberg correction.
- Procedural templates (Finding 5): each case's event sequence up to the sentence; for each judge and article group with ten or more decided cases, the share of cases on the judge's most common sequence, tested against 1,000 reassignments of cases among judges of the same court and article group, with Benjamini–Hochberg correction.
- Not run: prosecutor–judge pairing, because courts withhold most prosecutors' names.
Tables for follow-up
B1. EU listings that match a judge’s name and describe a judge (to be confirmed)
| Sanctions regime | Judge-court pairs | Listing names the same court |
|---|---|---|
| Repression in Russia (Reg. 2024/1485) | 6 | 5 |
| Ukraine territorial integrity (Reg. 269/2014) | 3 | 3 |
| Global human rights (Reg. 2020/1998) | 1 | 0 |
Counts only. Names are withheld until each match is confirmed against the Official Journal of the European Union (see the legal notice, section 5); they are shared with OVD-Info for review.
B2. The 21 judges recorded as presiding over cases with 25 or more convicted defendants
Source: the courts' case pages. Counts describe how often each judge presided; see the legal notice, section 4.
| # | Judge | Court | Convicted / acquitted | Custody extensions |
|---|---|---|---|---|
| 1 | Shelestov Andrey Nikolayevich | Central District Military Court | 48 / 0 | 21 |
| 2 | Kakadeyev Anton Igorevich | 2nd Western District Military Court | 46 / 0 | 10 |
| 3 | Kolesnikov Aleksey Valeryevich | Central District Military Court | 44 / 0 | 26 |
| 4 | Solomko Igor Ivanovich | Central District Military Court | 42 / 0 | 23 |
| 5 | Massin Yuri Aleksandrovich | 2nd Western District Military Court | 42 / 0 | 7 |
| 6 | Slepukhin Andrey Vasilyevich | Central District Military Court | 41 / 0 | 0 |
| 7 | Nikitchuk Igor Vladimirovich | 2nd Western District Military Court | 37 / 0 | 10 |
| 8 | Shishov Oleg Aleksandrovich | 2nd Western District Military Court | 34 / 0 | 22 |
| 9 | Pluzhnikov Andrey Viktorovich | 2nd Western District Military Court | 34 / 0 | 7 |
| 10 | Belkin Igor Viktorovich | Central District Military Court | 31 / 0 | 14 |
| 11 | Kiforenko Roman Ivanovich | 2nd Western District Military Court | 31 / 0 | 15 |
| 12 | Krasnov Vadim Vladimirovich | 2nd Western District Military Court | 31 / 0 | 8 |
| 13 | Chumakov Vitaly Sergeyevich | Central District Military Court | 30 / 0 | 23 |
| 14 | Ananyev Dmitry Sergeyevich | Central District Military Court | 30 / 0 | 12 |
| 15 | Kostin Igor Vladimirovich | Southern District Military Court | 30 / 0 | 0 |
| 16 | Raitsky Aleksandr Gennadyevich | Central District Military Court | 29 / 0 | 24 |
| 17 | Zhidkov Timur Georgiyevich | 2nd Western District Military Court | 27 / 0 | 7 |
| 18 | Mentov Vadim Gennadyevich | Central District Military Court | 27 / 0 | 2 |
| 19 | Torkunov Denis Mikhaylovich | Central District Military Court | 25 / 0 | 19 |
| 20 | Vladimirov Roman Viktorovich | 2nd Western District Military Court | 25 / 0 | 4 |
| 21 | Krylov Valentin Vladimirovich | Central District Military Court | 25 / 0 | 14 |
B3. 15 people under 280.4 part 2 not found in OVD-Info
15 people in 9 regions: 12 with a verdict, 1 with another final decision, 2 pending. Figures by region are not published because several regions have a single case. Courts, case numbers and links are shared with OVD-Info only.
B4. Judges sharing a full name with a person on Ukraine's sanctions register (to be confirmed)
| Court | Judges |
|---|---|
| Southern District Military Court | 29 |
| Other courts (one or two judges each) | 3 |
Counts only, and courts with fewer than three such judges are grouped: a shared full name is not proof of identity, so names are not published until each match is checked against the date of birth and decree.
B5. Judges in Finding 5
The judges identified by letter in Finding 5. A uniform recorded sequence describes entries on the case page, which the court's staff keep; it is not evidence of how any hearing was conducted, and does not imply that any judge acted improperly (legal notice, section 4).
| Presiding judge | District military court | Articles | |
|---|---|---|---|
| Judge A | Kakadeyev Anton Igorevich | 2nd Western | 205.2 |
| Judge B | Abasov Konstantin Aleksandrovich | Central | 205.2 |
| Judge C | Vladimirov Roman Viktorovich | 2nd Western | 205.2 |
| Judge D | Degtyarev Igor Petrovich | Southern | terrorism, treason, espionage |
| Judge E | Trifonov Nikita Ilich | Central | 205.2 |
| Judge A | Kakadeyev Anton Igorevich | 2nd Western | terrorism, treason, espionage |
What readers ask
Partly, and the report says so. Russian courts acquitted about 0.3% of defendants convicted or acquitted in 2022. The military courts here acquitted 0.04%, one person. Research on Russian criminal justice explains low rates by performance indicators for investigators and prosecutors, filtering before trial and judges' aversion to reversal. These cases sit at the far end of that pattern. What is specific to them is the charge structure: expression charged under a terrorism article, tried in military courts, above the penalty threshold at which ordinary ways out open (Findings 1 and 3).
The report does not claim that yet. The shares here are lower than national figures for the same five articles, but national tables count each article charged while this report counts whole cases, and for 205.2 this report already covers essentially every case in the country. A like-for-like test on 2023 decisions, single charges and defendants present in court is in preparation.
Not the judge alone. Investigators and prosecutors end many cases before they reach court, and dismissal at trial usually follows a motion by the prosecution or a party. The charge decides whether dismissal is possible at all; whether a defendant met the other conditions (a first offence, reparation, consent) is not on the case page. The report therefore measures how often dismissal was recorded where the charge allowed it, not how often a court refused it.
No. The cases are selected by article, not by motive. The military courts also try calls to violence, jihadist propaganda and armed attacks. Prosecutions of designated religious organisations, such as Jehovah's Witnesses under 282.2 and Hizb ut-Tahrir under 205.5, are a distinct group, and the Southern District Military Court also tries residents of occupied Crimea and captured Ukrainian servicemen (Finding 3). Where the report says "political", it means OVD-Info's classification.
Because the law assigns article 205.2, public calls to or justification of terrorism, to the district military courts (Code of Criminal Procedure, art. 31). The prosecutor chooses 205.2 over speech articles with lower penalties; how often cannot be measured from the case pages. Since 2009 the judges of military courts are not in military service, though many are former officers.
Many people charged under the anti-war articles have left Russia and are tried in their absence. They cannot meet some conditions for dismissal and their trials are short, which can lower dismissal rates and shorten records. They are not yet separated in this report; the next run will separate them.
A court order, not a verdict, applied to a person found not responsible at the time of the act, or who fell ill afterwards. It ends the case without a conviction but imposes treatment with no fixed end, reviewed by a court. A person who recovers from an illness that began after the act can be prosecuted. It is the commonest outcome other than conviction in these cases.
No. Judges are named because the courts' own public pages name them as presiding, in their official capacity. Counts describe how often a judge presided, not individual responsibility; a panel decides collectively. Uniform records (Finding 5) describe entries on the case page, not how hearings were run.
No. They are leads. An EU match means the name, and where stated the court, correspond to a listing that describes a judge; it must be checked against the Official Journal. Matches on Ukraine's register share a full name only and must be checked against the date of birth and the decree. Sanctions are preventive measures, not findings of guilt.
For the anti-war articles, nearly complete where courts publish names; less so under 280.4 and for cases registered in 2026, which have had less time to be documented. Where a court withholds a name, a match is possible only by docket number, so those cases cannot be checked (Finding 7).
The workbook behind the report is shared with OVD-Info only, because it contains case numbers and links to court pages that can identify defendants. The method, the definitions and every figure's source are in Annex A.
Write to ebrattin@traceoriginresearch.com. Any person named may respond; verified errors will be corrected and the correction dated (legal notice, section 7).
© 2026 Trace Origin LLC. Data: OVD-Info (used with permission); sudrf.ru; OpenSanctions; court directory Dataout / sudrfparser (CC BY-SA 4.0). Published by Trace Origin LLC, 17 October 2026; revised 27 October 2026, version 1.1. Corrections: ebrattin@traceoriginresearch.com.
Version 1.1, 27 October 2026. Title changed from “… Charged for Speaking Against the War”, because the cases are chosen by article and include calls to violence, armed attacks and treason; the scope is now stated in the opening. “Where the charge permits it” became “where the charge does not rule it out”. A sentence saying dismissal “sets these cases apart” was withdrawn: no like-for-like national comparison has been made. The 2022 national acquittal rate is now described as a reference point, not a like-for-like comparison. “963 cases” corrected to 963 registrations. The summary of Finding 2 now says custody is extended hundreds of times, rather than that defendants are held. Annex A now describes the waiting test as a closed-world query. A limit on appeals was added. Names of judges matching an unconfirmed EU listing were removed pending confirmation. No figure changed.