What You Would Have to Buy

Suppose you want me in prison.

I have not done anything, but treat that as a detail rather than an obstacle. You have decided I am a problem, and you have money and access to somebody who has access. You want a court in this country to put my name on a judgment. What do you actually have to buy?

I want the price list, item by item, because "here they can jail whoever they want" is a sentence I have been hearing since I was a child and it has never once arrived with one. It is always a mood. But the arithmetic can actually be done. I will start with the part of it that helps you, since an argument that opens with its own best case is not worth much.

It has been done here

Estonian judges have taken bribes, in Estonian courtrooms, with names and sentences attached.

Between 2003 and 2010 five bribery offences by sitting judges were registered in Estonia and all five judges were convicted. Ain Truu of the Tallinn City Court, 2003, two years, for a friendlier judgment in a case on his own desk. Ardi Šuvalov, a Harju County Court judge for fifteen years, convicted in January 2008 and given three years and six months for promising favourable rulings and an acquittal in exchange for money. Mihhail Komtšatnikov of Viru County Court, 2010, two years, for agreeing to take 100,000 kroons, roughly 6,400 euros, to release a murder defendant on bail. His colleague Jüri Sakkart, also 2010, two years and eight months, for taking a bribe and leaking surveillance material from a tax investigation, surveillance he had personally authorised. Vambola Olli, who ran the Võru courthouse, four years in November 2010, six months of it immediate and the rest suspended, for knowingly issuing unlawful decisions for acquaintances, plus an attempt at bribe-taking that never quite completed.

Prosecutors too. In 2015 an assistant prosecutor in the Northern District, Marina Kalistratova, was convicted of forging a suspect's signature on the consent form she needed to close a criminal case, then writing the termination order on top of the forgery. And this July, a district prosecutor in the Northern District, Eve Soostar, was convicted of official forgery for tampering with a live criminal file: she took the victim's signed refusal of plea proceedings out of the paper folder, destroyed it, and put in its place a consent form she had drafted herself. She got six months, suspended on a year's probation, and that judgment is not yet final. Over the last fifteen years five Estonian prosecutors have become criminal suspects, four were charged, two were convicted.

People also try from the outside. A Ministry of Justice study went out to all 410 Estonian judges and prosecutors in 2010 and 187 answered, of whom 4.3 per cent had been subject to attempted corruption in the previous eighteen months, against 0.3 per cent in Finland and 0.6 per cent in Sweden. Twenty-nine per cent of prosecutors and twenty-one per cent of judges reported some attempt to influence them. Of the eight people who said someone had actually approached them with corruption in mind, not one had reported it to the police.

So the fair starting position is that bribery on the Estonian bench is a known, documented, prosecuted category of crime with real names attached. Anyone who tells you the Estonian judiciary has never been dirty is selling something.

Which way the money was going

Read that list again and look at which way it points.

Truu sold a softer judgment. Šuvalov sold favourable rulings and an acquittal. Komtšatnikov sold bail for a man accused of murder. Olli quashed misdemeanour decisions for acquaintances. Kalistratova forged a signature in order to close a case, and Soostar forged a consent in order to push a plea agreement through. Almost every documented act of corruption or falsification in the Estonian justice system that I can find was somebody buying their way out, or an official cutting their own workload down. The one case pointing the other way, a constable firming up a witness statement (I come to her below), was about a 44-euro fine. Not one was an attempt to put an innocent person in prison.

That is seven cases running one way and one running the other. A pattern, not a law of nature, and worth exactly what a pattern that size is worth. It matters anyway, because the two crimes are not equally easy. Getting a guilty man released requires one decision from one official. Manufacturing a guilty man requires evidence that has to survive a second reader, and the price list from here on is mostly a list of second readers.

The file is the hard part

Start at the bottom of the chain, because that is where you would start too. You do not need a judge if you can arrange the evidence.

Look again at what happened to Soostar, because the interesting part of that case is not the crime, it is the detection. She swapped one page in one folder in an ordinary case, the kind nobody outside the room would ever read. What caught her was that the paper file and the digital copy did not say the same thing. The paper folder showed every victim consenting to a plea agreement. The digital copy was missing one of those consents, and in its place the victim's representative had written, in plain Estonian, that they did not agree to a plea agreement at all. The court noticed the contradiction and sent the file back. She was charged, and a judge convicted her.

No anti-tamper system caught her. What caught her was redundancy plus routine: two copies of one file living in two places, and a judge who read both. Multiply that one swapped page by an entire fabricated case and you are no longer forging a document, you are maintaining a second reality across a police system, a prosecution system, a court system and a defence lawyer who receives the file.

The version of this that people actually fear is smaller and involves no prosecutors at all: an ordinary officer writing something into a report that was never said. That happens here too, and the best-documented instance turns on 44 euros. In 2012 the head of a constable district in Pärnu, Eili Mesipuu, added a sentence to a witness statement above the signature already on it, to firm up a 44-euro mobile-phone traffic offence. The accused caught it. Katrin Pärn, the driver, read the file, noticed the line spacing and the overlapping text, and filed a criminal complaint; the witness was her mother. Mesipuu confessed, which is why no handwriting examination was needed, and Pärnu County Court convicted her of forgery in office.

Now the part that stops this being a story about a system that works. She was convicted by penalty order, a procedure with no hearing at all, and fined 780 euros, roughly two weeks of her pay. And in the weeks before her guilt was established she was moved onto her police pension, expressly so that she would not lose the entitlement that a conviction while still serving would have cost her. Her superior described her publicly as the good old salt of the earth. So the answer to "what happens to a policeman who falsifies a document here" is: he is prosecuted, by the quickest and least public route available, for about two weeks' wages, and somebody makes sure the pension survives. One absence points the same way. The Penal Code has a separate offence for removing or artificially creating evidence, and I could find no reported Estonian conviction of a police officer under it. Every one of these cases, from Mesipuu to Kalistratova to Soostar, was charged as forgery of a document.

What the defence lawyer receives is worth stating exactly, because disclosure is one of the places where the machinery is weaker than its reputation. Under the Code of Criminal Procedure the prosecution hands defence counsel a copy of the criminal file before trial, but it hands over the file as it has constituted it: the prosecution may first remove material it considers irrelevant to the case, and counsel who wants to know what was taken out has to ask. That is a real gap. It is also very far from the thing you need, which is a file nobody outside your control ever sees.

It is also not true, however often people say it, that illegally obtained evidence is thrown out in Estonia. There is no general rule to that effect. For ordinary evidence the Supreme Court weighs the breach against the seriousness of the offence. Only surveillance material carries an automatic rule: it counts as evidence only if the legal requirements were met, and there the court has no discretion to forgive. Ask for the legality of a surveillance operation to be checked and the court cannot refuse; if the accuracy of the written summary is in dispute, the recording itself has to be played, and a judgment that skips that analysis can be overturned for want of reasons.

This is not decorative. In April 2026 the Tallinn Circuit Court acquitted Kersti Kracht, a former ministerial adviser, on every count over 149 pages, and held that the prosecution had used intercepted material arbitrarily, cutting recorded conversations in a way that created the impression she had agreed to something unlawful. That finding did not come from a journalist or an activist. It came from the second-instance court, in writing, about the state's own prosecutors, in a case that had brought down a government five years earlier. The prosecution has taken it to the Supreme Court, so nothing there is settled.

Notice what that case is evidence of, though. Prosecutors in Estonia do cut corners, and when they do, the thing that catches them is a court reading the underlying recording.

What being caught actually undoes

Here is a question I did not think to ask until late. Suppose a fabrication is discovered. What happens to the conviction it produced?

The Code of Criminal Procedure has a reopening chapter, and two of its grounds are exactly on point: a judgment is reviewable if it rested on a forged document or on artificially created evidence, or if an official of the pre-trial investigation or a prosecutor committed a crime during the proceedings that may have affected the outcome. Read the conditions attached to both, though. Each requires the forgery or the crime to have been established by a separate court judgment that has itself become final. You cannot reopen your case by arguing your evidence was faked. You can reopen it once somebody has been convicted of faking it.

That is a high gate, and it makes the Soostar conviction more than a scandal, because a conviction for official forgery is, once it becomes final, exactly the separate judgment the reopening chapter demands. It also means the ordinary sequence is very slow. Estonia's one well-documented reopening, decided by the Supreme Court sitting as a full bench of eighteen in April 2012, concerned Henrik Antsov, whose conviction had rested on a witness whose account a later judgment contradicted. The court struck down part of the reopening provision as unconstitutional in order to get to him, quashed the relevant part of his conviction and resentenced him. He had already been released on parole two and a half years earlier. The correction arrived after he had served the time.

And here is a negative that cuts against my own argument: I could find no Estonian case in which a conviction was overturned because the state was found to have fabricated the evidence. The near misses all sit before finality, where a court threw the material out in time. Catching it in time is better than fixing it afterwards. It also leaves the afterwards untested.

The judge

Say you get past all of that and go for the judge directly.

First you have to know which judge. You cannot, in the ordinary case, because the Courts Act requires the annual work-distribution plan to allocate every incoming case among judges on the principle of randomness, and it can only be changed mid-year for good cause. The plan is public. Two exceptions are worth knowing, and a sceptic should have both: during a state of emergency or war a court chairman may rewrite the plan by order, and the court reform adopted in June 2026 creates a presidium inside each court which takes the work-distribution plan away from the plenary of all its judges. Concentrating case allocation in a small body is the precise mechanism by which cases were steered in Poland and Hungary. In ordinary peacetime today you are still buying a lottery ticket rather than a person.

Then there is the person. An Estonian judge is appointed for life and can be removed from office only by a court judgment. Nobody staffs the bench alone either: the Supreme Court en banc proposes exactly one name per vacancy and the President's only power is to say no. A first-instance judge's salary is set by statute at 0.65 of the highest state-servant rate, which in April 2025 came to 5,858 euros gross a month, and the chairman's supplement is a formula based on how many judges sit in the court rather than something anyone negotiates.

One fact belongs right here, and for a reader who distrusts the security service it outweighs everything else in this section. Every candidate for judicial office in Estonia undergoes a security check by the internal security service before appointment. If you believe KAPO effectively decides who gets to be a judge, that single provision looks like the answer to all of it. What it actually gives the service is a veto at the entrance, before anyone has a case. It does not tell the service which judge gets your file, and it does not survive the judge's appointment for life. It has also not stopped Estonian courts from acquitting people the service spent years investigating, which is most of the second half of this essay. It is a real lever and it is at the wrong end of the process to deliver you a verdict.

Now price the risk. To charge a sitting county or circuit court judge with anything at all, the Prosecutor General has to petition the Supreme Court en banc, the en banc has to propose it to the President, and the President has to consent. That gate protects judges from arbitrary prosecution, and it works: in October 2020 the en banc refused to forward such a request because the accusation was not clear and precise enough to put before the head of state. It cuts both ways, obviously. The same gate means a genuinely corrupt judge is screened, in the first instance, by other judges.

But it is a gate on prosecution, not on observation. Šuvalov, the judge who got three and a half years, was caught by the internal security service in a controlled operation authorised by a court president. He took his conviction to Strasbourg arguing his trial had been unfair, and lost.

Here is the number that tells you where Estonia is now. The last judge this country put on trial for corruption was Eveli Vavrenjuk of Tartu County Court, and the total value of everything the prosecution alleged she received, across eleven years, was 1,514 euros and 99 cents: twelve hours of sporadic free legal advice, five law books for her student son, and help buying a phone at a discount. She was acquitted, the acquittal was upheld in 2024, and the state was ordered to pay her and five co-accused over 263,000 euros.

An investor would price that differently from a partisan. In 2003 a bribe bought a judgment. In 2021 the state opened a criminal case against a judge over a sum a Tallinn law firm bills in two days, spent years on it, and lost. I have found no publicly reported conviction of an Estonian judge for corruption since 2010, and I want to be careful about what that sentence can carry: an absence of reported cases is not proof of an absence of cases, and in a country of 1.36 million with 249 judicial posts it is exactly what undetected corruption would also look like. The argument I can actually make is a probabilistic one. When the threshold for opening a judicial corruption case is fifteen hundred euros of favours spread over a decade, a large undetected market in bought verdicts is unlikely, because a market that size would have to be invisible to a state that goes after sums that small.

And then the court above that one

Suppose you buy your judgment anyway. You have bought one document, and that document is now going to be read by strangers.

Every criminal judgment from a contested trial can be appealed on the facts to a circuit court and then on the law to the Supreme Court. The instances do disagree with each other. Kalle Laanet, a former Minister of Justice, was convicted of fraud by the county court in February 2026 and acquitted by the circuit court in May, with the state ordered to pay him 36,548 euros and 40 cents. Disagreement runs the other way too, which the essay would be cheating to hide: in the Porto Franco affair the county court acquitted the businessman Hillar Teder, the Centre Party's secretary-general and the party itself, the circuit court reversed and convicted, and the Supreme Court made the conviction final in February 2025 with a one million euro aggregate penalty against a party that had been in government. A first-instance acquittal is a stage in the process, and both sides know it.

Above all of that sits Strasbourg, and I want to describe it accurately rather than triumphantly, because it is a weaker backstop than people imagine. Estonia has had 66 cases transmitted to the Council of Europe's Committee of Ministers since it ratified the Convention and has closed 64 of them; at the end of 2025 two remained pending, both under ordinary supervision, none under the enhanced procedure used for serious or structural problems. In 2024 the Court delivered no judgments against Estonia at all.

The most on-point case Estonia has lost is Dorozhko and Pozharskiy, where the applicants complained that their trial judge was not impartial because her husband had headed the police team investigating their case. Strasbourg agreed, holding that there were ascertainable facts raising objectively justified doubts about the presiding judge's impartiality. Here is the part that matters: the convictions were not quashed. The Estonian Supreme Court examined a request to reopen the proceedings on the merits and held that the finding did not call the outcome into question, the applicants received 1,500 euros, and the Committee of Ministers closed the file in 2010. That is what winning in Strasbourg often looks like: a declaration, a small payment, and the judgment against you still standing.

Both applicants in that case were Russian-speakers, which is worth saying plainly to the people most likely to disbelieve this essay. Russian-speaking applicants have beaten the Estonian state in Strasbourg repeatedly, in Shchiglitsov, in Missenjov, in Kochetkov, in Mikolenko, in Andreyev, in Korobov over the Bronze Night violence, in Nikitin over prison conditions. Strasbourg has ruled against this state for people with your surname more than once. Its rulings still do not usually give you your life back.

The comparison you are actually making

None of this lands unless I say out loud what the sentence "they can jail whoever they want" is really about, because it is usually not about Estonia. It is imported.

In Russia last year the courts convicted 468 people of treason and espionage and acquitted nobody. Not a low number of acquittals. None. Across that entire category of case there has been one acquittal since 1997. In ordinary Russian criminal proceedings the chance of acquittal runs at about 0.15 per cent, one in 676. A person convicted in a Russian court today also has no route to Strasbourg at all, because Russia stopped being a party to the Convention in September 2022.

A low acquittal rate on its own does not carry much weight, because Estonia's own headline rate is about one per cent and Germany's is under four, and the number tells you mostly about how cases are filtered before trial. What distinguishes the Russian picture is the combination: near-zero overall, exactly zero in the political category, and no external court left to appeal to.

The clearest Russian counter-example proves the same point. When police planted drugs on the journalist Ivan Golunov in 2019, five officers were eventually convicted of fabricating the case and their commander got twelve years. It took a national press revolt, three newspapers running the same front page, and a level of public attention no ordinary defendant will ever attract. Read it for the price of detection: there it took a national firestorm, here it took a judge reading a file.

It is also worth knowing that Estonia never purged the judges it inherited. There was no lustration of the Soviet bench. The People's Court judges of the Estonian SSR, seventy-nine of them in 1987, were replaced as their terms expired, through open competition, with a written oath of conscience in place of a security screening. The institution was rebuilt by turnover rather than by decree, which is slower and less satisfying and, judging by where the country sits now, worked.

Where the argument runs out

If the essay stopped there it would be a brochure, so here is the other column, and it is not small.

First, most Estonian criminal cases are never tested by anybody. In 2025 the county courts decided about 3,180 criminal cases, and 78 per cent of them ended in a plea agreement. Seven per cent, 228 cases, went through the full adversarial procedure where evidence is examined in open court and the defence can attack it. Every safeguard I have just described bites only inside that seven per cent, and a plea judgment cannot be appealed on the facts at all. So the claim has to be stated conditionally or not at all: Estonian courts test evidence when a defendant makes them, and in four cases out of five nobody does.

Which raises the obvious question of why. Part of the answer is that nobody can be made to sign: the accused, his lawyer and the prosecutor all have to. But a signature is not an admission of anything, as I come to below, and the rest of the answer is that resistance is priced. The median plea took 23 days in 2025 against 203 days for a full trial, and the longest first-instance proceeding to end that year ran 2,652 days, and it was a plea, which is its own comment on how much of the delay a defendant controls. At the end of 2025, 388 of the 1,621 people held in Estonian prisons and detention houses were remand prisoners who had not been convicted of anything, and that number rose over the year while the convicted population fell. The Council of Europe's anti-torture committee, after visiting in 2023, called it unacceptable that Estonian remand prisoners are still locked in their cells 23 hours a day for long periods. The Supreme Court has held that a person who cannot get adequate legal aid because of the fees the state pays its appointed lawyers may have his right to a defence violated. Nobody has to fabricate anything to get a signature out of that arrangement. My own thought experiment answers itself at this point: to break someone here you would not need to buy a thing. You would only need to make fighting expensive.

There is a bridge available here and I will take it, and then show you where its plank is missing. The scenario people actually fear, a case invented against a specific person for political reasons, is by definition a contested case, and contested cases are the ones that run the full machinery. Peterson, Seredenko, Bessedin, Kracht, Teder, Laanet, Reps, Vavrenjuk: general procedure in every one of them, with reasoned judgments pronounced publicly and appeals. But the plank is missing where it hurts most, because every one of Estonia's first twenty espionage and treason convictions ended in a settlement, which means that in the category this readership distrusts most, the state's evidence has almost never been tested in open court by anybody.

Second, the decisions that can end your life here fastest do not start in a criminal court at all. A residence permit is not a criminal charge. Estonia's own Chancellor of Justice reports that the police have refused to extend, refused to grant, or moved to revoke the residence permits of Russian citizens already living in Estonia, in cases where nothing about the person had changed and they had done nothing here to be held against them. In January 2024 the Police and Border Guard Board announced it would not renew the permit of the head of the Estonian Orthodox Church of the Moscow Patriarchate and required him to leave. In January 2026 the Ministry of the Interior imposed entry bans on 261 Russian combatants in one batch, a ministerial decision. When the former diplomat Clyde Kull challenged the refusal to renew his security clearance, the Supreme Court upheld the refusal while noting that the security service had erred by interviewing him in the administrative proceedings while he was simultaneously detained as a criminal suspect, an error it found had not affected the outcome. All of this is reviewable in the administrative courts, which is a real protection. It is also decided first by an official, on a standard of proof that is not "beyond reasonable doubt", on a file that can be partly classified. And the review has a shape worth knowing before you rely on it. There is no special security tribunal; you go to the ordinary administrative court. What is special is that the court must decide separately, and give separate reasons, on closing the proceedings, on removing you and your representative from the examination of the evidence, and on refusing you sight of the file, with each decision appealable on its own. The Supreme Court has held that removing a party from his own case requires an interest in secrecy that is clearly and unhesitatingly weightier than his right to be present, and that a court has to reason its way there case by case rather than point at the statutory ground and stop there. All of which still leaves you appealing a decision whose factual basis you have never seen, and in the administrative track there is no automatic exclusion rule at all. When the Supreme Court reviewed its own published practice on closed administrative hearings, it found not a single state-secret dispute in the sample. You can read that as vanishingly rare or as invisible. I lean to the first, and I have to admit the data cannot separate them.

Third, the numbers I have been quoting about how clean this country is are mostly perception, and perception is not incidence. Estonians report far less corruption around them than the European average. On the questions that ask about experience rather than impression, Estonia is ordinary: 12 per cent of Estonians say they personally know someone who takes bribes, against 10 per cent across the EU. And one figure belongs here precisely because it damages the argument. Among Estonians who had contact with the courts in the past twelve months, 10 per cent said someone asked or expected a gift, a favour or extra money, against 4 per cent across the EU. That sub-sample is small and the survey warns against reading national figures off it, and I still think a person making my argument does not get to print the flattering numbers and skip that one.

Fourth, some of the state's own housekeeping is bad. The European Commission has now told Estonia in three successive annual reports that the body governing its courts does not meet European standards, and Estonia passed its court-administration reform in June 2026 without fixing it. Judges' pay indexation has been frozen since 2024 and 55 judges sued the state over it and lost at first instance. The courts' case-management system is legally owned by the ministry rather than by the judiciary. Courts almost never refuse a surveillance warrant: in 2025 they fully rejected 13 applications while granting 459, and the prosecution issued a further 312 authorisations on its own signature. A refusal rate of under three per cent is what a rubber stamp looks like from the outside, and I cannot tell you from published data whether it means the applications are good or the scrutiny is thin.

And 59 per cent of Estonians rate the independence of their courts as good, which the Commission classes as merely average, while domestic trust in the court system has fallen from 65 per cent to 62. Neither figure is broken down by native language. What Russian-speakers in particular think is therefore unmeasured, and the national average does not speak for them.

The cases you are really asking about

Nobody writes to me about bribery in a bankruptcy case. The letters are about treason, and about people with Russian names.

Take the shape of those cases seriously before taking their politics seriously. Estonia's first twenty convictions for espionage and treason, from 2009 to late 2019, share one feature: every single one was concluded by settlement procedure, an agreement signed by the accused, his lawyer and the prosecutor.

I had that wrong when I started writing this, and the correction runs against my own argument, so it goes here rather than in a footnote. A settlement in Estonia is not a confession. The Code of Criminal Procedure nowhere makes an admission of guilt a condition of reaching one. What the judge has to establish is narrower: that the accused understood the agreement, consents to it, and is expressing his true will. The Supreme Court held in 2005 that a settlement is available even where the accused denied his guilt at the earlier stages of the proceedings, and upheld a conviction in exactly that situation. The procedure in which an Estonian court actually asks a defendant whether he pleads guilty is the abbreviated procedure, which is a different thing and which the twenty did not use.

So "all twenty settled" and "all twenty admitted it" are two different claims, and only the first is established. An admission is documented for two of them. Herman Simm's prosecutors said he pleaded guilty and cooperated, which is why they did not seek the maximum. Deniss Metsavas told his story on camera after his conviction. For the other eighteen there is no public record either way, because the hearings were closed.

Two further features of that procedure belong here, and both cut against the state rather than for it. In a settlement an Estonian court cannot acquit: the only judgment available to it is conviction on the agreed terms. And once the accused has confirmed the agreement in court, the road to appeal is almost entirely shut. What the court can do, and must, is refuse the agreement and send the file back if it doubts that the act happened, that this person committed it, or that it amounts to a crime at all. That is a judge reading a file, not a defence testing evidence in open court, and it is the thinnest check anywhere in this essay.

Who were they? Half were Estonian citizens. Two were retired officers of the internal security service and one was arrested while still serving in it. The largest case of them all was Herman Simm's. An Estonian who ran security at the Ministry of Defence, he got twelve years and six months. If this law were an instrument for punishing Russians, somebody forgot to tell the people applying it.

The contested cases are the ones where the machinery had to work. Sergei Seredenko, whom Russian state media file under political prisoners, admitted nothing. His hearings were held behind closed doors, the judgment was pronounced publicly, and the sentence he received was longer than the prosecution had asked for, which is not what a court taking dictation does. The Supreme Court upheld his conviction and simultaneously used his case to set limits on the security service's use of intelligence as evidence, holding that as a rule the accused and the defence must see the evidence against them, and it ruled partly in his favour on confiscation. In the Aivo Peterson case, which I wrote about when the verdict came, the county court refused to take the indictment whole, rejecting the civil-defence allegation as treason and finding part of the rest unproven; it granted him bail of 150,000 euros during proceedings, and the circuit court that raised his sentence to sixteen years also ordered the state to pay him 1,885 euros in costs. In the case of Gerli Mutso, convicted of working for Chinese military intelligence, the Supreme Court annulled part of the conviction on strict legality grounds, because her cooperation before a 2019 amendment had not been punishable when it happened.

And now the version of the fear that I think is actually correct, at full strength:

Nobody has to buy anything. The service decides you are dangerous on material gathered under a law that needs no judge and allows the service never to tell you. It hands a conclusion, not the source, to a prosecutor. You are charged not with an act but with a relationship. Your hearing is closed in substantial part on state-secret grounds. Your lawyer gets a summary. You sit in remand while it proceeds. At the end a judge whom that same service vetted before appointment, working in a language you speak second, convicts you, lawfully and in a reasoned public judgment that an appeal court reviews and partly rewrites. Nothing in that sequence had to be fabricated.

Almost every element of that is built from documented Estonian fact. In these cases a central piece of evidence can be an intelligence summary produced by the security service, a document of conclusions where much of the underlying material never enters the case file because it is classified; the Supreme Court said so openly in Seredenko's case. A defence advocate, a former Supreme Court justice, the chief state prosecutor and a former head of the security service sat down for a published panel discussion in Juridica in 2024, and the first two put the equality-of-arms problem plainly: the defence receives conclusions rather than data and cannot test what it cannot see. That is the Estonian legal profession arguing with itself, in its own journal. The lesser covert measures under the Security Authorities Act are authorised by the head of the service rather than a judge, and if a year passes after declassification and the grounds still hold, that same official may decide you will never be told at all, except for the most intrusive measures, where an administrative court judge has to approve the silence. Estonia's own Chancellor of Justice has written that current law leaves higher courts very few opportunities to assess this track at all.

Strasbourg has already ruled against Estonia on this. In Leas v. Estonia, decided in 2012, the mayor of Kihnu was convicted of soliciting bribes on the strength of a surveillance file the defence was not allowed to see in full, and the Court found a violation of the right to a fair trial.

So what is my answer to that paragraph? Only this, and I want it measured. A court that had been handed its verdict would not refuse the indictment it was handed, would not set bail, would not order the state to pay the defendant's costs, would not hand down a sentence the prosecutor did not ask for, and would not spend a judgment writing a rule that constrains the service for the next defendant. Those are not the moves of a rubber stamp. They are also not proof that the underlying evidence in a closed session was sound, and I cannot give you that proof, because neither of us has seen it.

Two more things belong here. The first is a hole in my own argument, and it is not a small one. Not one person convicted of treason or espionage in Estonia has ever obtained a judgment or a decision from the European Court of Human Rights. The external safeguard I keep pointing to has never been tested on the category of case where suspicion is highest, in part because almost all of those defendants pleaded guilty and a settlement closes that road. The most serious international criticism, a 110-page opinion on the Peterson case by the emeritus professor Douwe Korff, exists and deserves reading, and its provenance deserves stating in the same breath: it was commissioned by a Geneva NGO on whose board Korff sits, and written before the domestic judgment it criticises.

The second is what it costs to be innocent here, and for that I would rather quote the other side than myself. Oliver Nääs is a defence advocate, one of the lawyers who sat on that Juridica panel, and in an interview published by the Estonian Bar Association he says both halves of this essay better than I have. Arbitrariness by investigators, he says, still occurs in criminal proceedings, not in every case, but you have to be ready for it in any case, and then: fortunately the courts do pay attention to deliberate procedural violations and abuses by the authorities. That is my thesis, from the person paid to disbelieve it.

His complaint is the other thing. There is, he says, no culture of apologising anywhere in the justice system: the state can put anyone on trial, whatever then happens to them it simply turns the page and moves on, while the person is left picking up the pieces of his life. The possibilities for compensating that damage he calls practically nonexistent. He asks whether what we practise is punishment by means of procedure. And he lists, by name, the defence lawyers acquitted in recent years, Tuul, Namm, Särgava, and wonders aloud whether they were prosecuted for what they do for a living. He points at Valvo Semilarski, named as a bribe-taking official at a prosecution press conference, whose case was then terminated without ever being sent to a court, and who had to sue to get the statements taken back.

That is not a marginal grievance. The European Commission opened an infringement procedure against Estonia over its transposition of the presumption of innocence directive, and the deficiency it named was public statements referring to a person as guilty prematurely or wrongly, together with the remedies available when that happens. Kracht was cleared after five years and two instances. Vavrenjuk was cleared after losing years of a judicial career. Laanet was cleared on appeal, after a conviction. And when Küllike Namm, acquitted in a case the state had built, claimed compensation for an unfounded charge, the county court refused it. Acquittal buys your legal costs. It does not buy the years, and this country has not built much of a mechanism for the difference.

What I actually think

Buying an Estonian judge is not impossible. It is unaffordable and, more to the point, useless.

Unaffordable because of what the seller has to stake. A judge here has a job for life at roughly three times the average wage, and to sell you a judgment he puts that, his pension, his profession and his liberty against a bribe that Estonian courts, in convicted cases, have historically measured in hundreds or low thousands of euros. Useless because of what you would get. One document, in a system that assigns cases by lot, records its own hearings, keeps the file in two places, hands the defence a copy, cannot refuse to check a surveillance permit when asked, and sends the result to two more courts and then, if you are still unhappy, to a court in France that has ruled against this state for Russian-speaking applicants more than once. You would be buying a delay, at the price of everything you have.

That is my answer, and I hold it: the story about cases being stitched together to order in this country is false, and the five convicted judges of 2003 to 2010 support that rather than undermining it, because we know their names, we know their sentences, and the going rate since then has collapsed to a level the state prosecutes and loses at.

But it does not cover what people are actually afraid of, and the gap is where this piece has to end. Everything above concerns criminal courts. If a state wanted to make life impossible for a Russian-speaking person in Estonia, it would need neither a judge nor a forged page. It would need a residence permit that is not extended, a clearance that is not renewed, an entry ban signed by a minister, a file with a classified annex, a plea agreement offered to a frightened man who cannot afford to keep paying his lawyer, and a proceeding that takes four years and ends in nothing, having already taken the four years. None of that requires anyone to be bought. All of it is legal, and some of it may even be right.

So keep the suspicion. Point it at the door that is actually open.