The Freedom Institute grades how countries censor, measures what privacy products actually do with your data, and publishes the evidence behind every score. We build the testing tools in the open so anyone can repeat the work.
First edition The Privacy Product Index opens for public methodology review Sep 2026If you can be identified, you can be targeted; most people have no way to find out which tools actually stop that.
A VPN provider represents that it retains no logs, a messaging platform represents that its users cannot be identified, and a cloud service represents that deleted data has been deleted. Each of these is a representation made by the party holding the commercial interest in it being believed, and almost none of them arrive with evidence that a reader can verify independently. The same gap runs through state conduct: a government restricts a protocol on a Tuesday and restores it on a Friday, and the record of what happened has usually disappeared before anyone documents it.
We publish country censorship grades and a Privacy Product Index that separate what a provider says from what we observed, what an auditor confirmed, and what nobody knows yet. The unknowns get printed too.
Research that nobody can act on is a publication, not a public good. Every grade and every index entry is written so that these three can make a decision with it.
They need to know what a device gives up when it is inspected and what a provider hands over when it is asked, before the trip rather than after it.
They coordinate in real time under conditions where group membership is the sensitive fact and the connection itself may be the evidence against them.
They are picking between two services on a Sunday afternoon and deserve a straight answer on which one collects less, keeps it for a shorter time, and lets them take it back.
Advice that ignores who is looking for you is guesswork. Pick the situation closest to yours and the country tier you operate in; the panel rewrites the order of what to check first.
Guidance shape only. The published version links each item to the test that produced it.
Each country is scored on four axes. Move the weights and every grade in the table recomputes, because a scoring choice is an argument and you should be able to test ours. Open any row for the axis detail, the shutdown history, and the evidence log behind it.
| # | Country▾ | Grade▾ | Blocking▾ | Shutdowns 12M▾ | Trend | Circumvention▾ | Evidence▾ | Confidence▾ |
|---|
Providers write their own privacy claims and grade their own homework. Independent verification is rare, expensive, and almost never repeated after launch, so a claim that was true at audit can quietly stop being true and nobody finds out.
A throttle lasts six hours; a block on one protocol lasts a weekend. By the time a report is written the traffic has recovered and the evidence is gone. Without continuous measurement, the worst behaviour is the behaviour that leaves no trace.
A journalist in Tehran, an organiser in Yangon, and a researcher in Minsk all need the same thing: which of these five tools will not give me up, and how do you know. The honest answer today is usually a forum thread and a review paid for by affiliate revenue.
Country scores cover blocking, shutdowns, restrictions on circumvention, and punishment for online expression. Where our coverage is thin, the confidence bar says so rather than rounding the gap into a number.
The Privacy Product Index compares tools inside a category on data collection, retention, provider access, tracking, privacy defaults, and user control; each cell is labelled claimed, observed, audited, or unknown.
We ship the diagnostics we test with, work with product teams on the fixes our testing surfaces, and turn findings into guidance people can act on and briefings regulators can use.
| Entry▾ | Category▾ | Composite▾ | Data collected | Retention | Provider access | Third-party tracking | Verification▾ |
|---|
The signature series takes the three promises the industry repeats most often, no logs, anonymous, and deleted, and tests each one against what the software and the servers actually do. Everything after it follows the same rule: publish the method, publish the failures, publish the unknowns.
Our largest funder does not get a soft review; we wrote that into the structure before we wrote a word of research.
UR would be the Institute's primary supporter and its founding organiser, with funding and in-kind contributions disclosed in full. That relationship creates the obvious conflict, because UR builds the category of product the Institute exists to assess. Independence that depends on goodwill is not independence, so control of methodology and publication sits with a majority-independent board and autonomous research leadership, and any assessment touching UR or anything the Institute builds is led by external reviewers.
Every funder, grant, and in-kind contribution published with amounts and dates.
Majority-independent board; research leadership holds final say on what publishes.
External reviewers lead any assessment touching a funder or an Institute product.
No sponsored placement, no affiliate revenue, no sponsor veto over a finding.
UR is a global institute for privacy and encryption products. Its network is contributed rather than rented: more than 80,000 members supply bandwidth that anonymises traffic for people who cannot route around their own government. UR would organise and fund the Freedom Institute, and would also sit inside its scope of assessment.
Organiser and primary supporter
Commercial privacy infrastructure concentrates trust in one operator and one jurisdiction; a member-contributed network spreads it across the people using it. That design is the reason UR can serve places where a single corporate exit node would be blocked or seized within a week, and it is also the reason UR's own claims need testing by someone who does not report to UR.
The Institute publishes the test results either way. A finding against UR is the proof that the structure works.
Structured review of corporate disclosures on privacy and expression.
Research processCriteria-based recommendations with rules against provider influence.
privacyguides.orgAn organisation that asks providers to show their evidence should show its own. All photography is sourced from Wikimedia Commons under public domain, CC0, or Creative Commons Attribution terms, resized but not otherwise altered, and credited below and on hover.