Expression is the default
Restrictions on lawful speech should be specific, proportionate and appealable. Blanket filtering of a protocol, a domain or an idea is none of those things.
Independent · Non-partisan · Reader-funded
LibreNet documents online censorship, defends lawful free expression, and publishes the evidence that keeps the network open. We work in the open, we publish our sources, and we take no money from platforms or governments.
For most of human history, the cost of publishing an opinion was measured in money, paper and risk. Today it is measured in milliseconds — and controlled by a small number of intermediaries: platforms, carriers, registries and the firewalls that sit between them. When a lawful voice is silenced online, it is rarely silenced by a court. It is silenced by a filter list, a takedown template, or an unannounced block that nobody is allowed to appeal.
LibreNet exists to make those decisions visible. We measure what is blocked, we record who ordered it, and we publish the results in a form that journalists, lawyers and legislators can actually use. Where censorship is unlawful, we support the people challenging it. Where it is merely opaque, we insist on transparency.
We are not a platform and we host no user content. We are a small research team, a public archive, and a set of tools — and we believe that a network where ideas travel freely is more stable, more prosperous and more peaceful than one where they do not.
Four commitments we apply to our own work before we ask anyone else to.
Restrictions on lawful speech should be specific, proportionate and appealable. Blanket filtering of a protocol, a domain or an idea is none of those things.
A person cannot speak freely while being watched. Anonymity and end-to-end encryption are not loopholes to be closed — they are the conditions of dissent.
Blocklists, removals and network interference should be documented and published. A restriction nobody can inspect is a restriction nobody can challenge.
We publish measurements, methodology and raw data. Conclusions follow the record, and we correct our own work in public when the record changes.
Selected publications from the current cycle. Full datasets are released with each report.
We tracked the propagation of transport-level blocklists across 27 national networks over eleven months. Nearly three quarters of the entries we observed appeared in at least two jurisdictions before any public notice, suggesting shared vendor feeds rather than independent national decisions.
An audit of 6,300 removal requests submitted through automated channels. We estimate that between 8% and 14% targeted content that was lawful in the requesting jurisdiction, and that fewer than 2% of affected publishers were told why their material disappeared.
Seven years of longitudinal measurements of reachability, latency and interference across 1.2 million endpoints. The index, the collection code and the anomaly log are published together so that other researchers can reproduce every figure.
A comparative review of fourteen legal frameworks and the technical measures applied under them. In nine of the fourteen, the infrastructure in use was capable of blocking far more than the law authorised, with no published limit on its use.
LibreNet is an independent research initiative founded in 2019. We are funded by unrestricted grants from foundations that support civil liberties and open science; we accept no funding from advertising, platform operators or any government. Every report names its funders, and our accounts are published annually.
Our team is distributed across five countries and works remotely by necessity as well as by preference. We publish in English, and we release our methodology in full so that our findings can be checked by anyone with an internet connection — which is, after all, the point.
This site is deliberately self-contained: no third-party scripts, no analytics, no fonts or assets loaded from other domains. What you read here is served from one server, and nothing about your visit leaves it.