The privacy penalty

Read naively, the gradient is a paradox. The man who marched in the square gets 5 years; the woman whose private messages were intercepted gets more; the one whose phone was unlocked and read, 10. States are supposed to punish the public challenge hardest. The paradox resolves when you notice what each evidence channel is allowed to prove. A public act can only support a public-act charge — "discrediting," a protest article — and those statutes are calibrated in single digits. Evidence extracted from private space feeds the heavy chapters of the code: conspiracy, treason, terrorism financing, "confidential cooperation." The investigative instrument comes first; the charge follows from the instrument, and the sentence follows from the charge. Which is the actual finding: the intrusiveness of the investigation predicts the sentence better than the visibility of the act. Tell us only how the evidence was gathered — a street camera, an informant, a forensic phone extraction — and you can roughly price the sentence before the charge is filed. And where the state found no act at all and manufactured one — planted cartridges, "explosives" in a bag — the median is 9 years, near the very top of the scale. A state that has already decided on a heavy sentence procures evidence to match.

Every documented sentenced case worldwide with a classified basis (898 cases across 8 countries; bases with at least 20 sentenced cases shown). Red bars are evidence the state extracted from private space — or manufactured; grey bars are public acts.

Russia only — the largest single-country sample (465 sentenced cases), so the pattern is not an artifact of comparing harsh-sentencing countries against lenient ones. The same courts price a public post at 5 years and the contents of a phone at 11.5.

• §These are cases documented in our database — the casework of Viasna, Memorial, OVD-Info and regional monitors across 8 countries — not a random sample of prosecutions. The claim is about documented political cases, and it is checkable case by case in the receipts.
• §This is a correlation across cases, not a sentencing rule. The mechanism runs through charging: private-channel evidence enables graver statutes (treason, terrorism, "extremist organization"), while public acts feed speech and public-order statutes with lower tariffs. The gradient measures that routing — which is the finding, not a confound.
• §Cross-country composition could inflate a global gradient, which is why the second chart repeats the analysis inside Russia alone, where it holds at least as strongly (5 → 11.5 years). In Belarus the gradient is flatter: its device-evidence cases sit near the public-act median, and its heaviest sentences concentrate in fabrication and informant cases instead.
• §Only sentenced cases are counted — people in pretrial detention are excluded, and cases that resolve fastest tend to be the gravest. That selection affects levels, not the comparison between bases, which are all drawn from the same sentenced pool.
• §The basis of each case is an editorial classification from cited evidence, with a confidence score and a verbatim quote per case (shown in the receipts). 585 sentenced cases have no classifiable basis and are excluded; bases with fewer than 20 sentenced cases are not charted.

Every documented case behind this finding, each linking to its profile and primary source. 144 carry a verbatim quote of what the prosecution rested on.