Protect people from harm without giving the state ownership of truth.
Legal traditions differ, but the core problem is worldwide: rules for real harms can become tools against criticism, minority viewpoints, journalism, history, science, and ordinary disagreement.
Editorial research synthesisSafety and institutional powerWorldwide comparative legal and human-rights frameworkUpdated 2026-09-02
Direct answer
A legitimate restriction should be clear in law, pursue a recognized and specific aim, require intent where appropriate, show a close connection to likely harm, use the least restrictive effective measure, provide independent review, and preserve the freedom to seek and receive information. Vague offense, discomfort, disagreement, or state embarrassment are not enough.
Key points
Freedom of expression protects offensive and disturbing ideas, not only agreeable ones.
True threats, direct exploitation, fraud, and incitement can be regulated without making entire topics forbidden.
Context, speaker, intent, content, reach, and likelihood prevent labels from replacing analysis.
Private platforms retain editorial interests, but consequential enforcement still needs transparency and remedy.
The right protects the process of seeking as well as speaking
Expression law often focuses on the speaker, but cognitive liberty also protects the listener, reader, researcher, and person still deciding what they believe. A rule that leaves speech formally legal while blocking discovery, criminalizing receipt, or identifying every reader can damage the same underlying freedom.
International human-rights principles recognize seeking, receiving, and imparting information across borders. Restrictions remain possible, but they must be treated as exceptions. The state should not reverse that relationship by requiring citizens to prove that every controversial question is safe before access is allowed.
The harm category must be operational, not rhetorical
“Harmful speech” can refer to very different phenomena: a true threat directed at a person, intentional incitement likely to produce violence, defamation, exploitation, fraud, harassment, political propaganda, inaccurate medical claims, offensive art, or criticism of institutions. Applying one automated rule to all of them guarantees error.
A rights-preserving system defines the protected interest, the prohibited act, the required mental state, the causal connection, and the available remedy. It distinguishes disagreement from threat, analysis from advocacy, and advocacy from operational coordination. The more severe the penalty or prior restraint, the stronger the evidence and review should be.
Use a six-factor threshold before coercive restriction
ContextWhat social conditions and vulnerabilities make harm more or less likely?
SpeakerWhat authority, influence, or relationship does the speaker hold?
IntentWas harm advocated deliberately, or is the material analytical, satirical, or mistaken?
Content and formIs there a direct call, threat, or operational instruction?
ExtentWhat is the reach, repetition, targeting, and distribution method?
LikelihoodHow probable and proximate is the alleged harm?
These factors do not mechanically decide every case. They prevent a broad label from doing all the legal work.
A worldwide due-process floor
Publish the rule and define its terms before enforcement.
Require a reasoned decision tied to specific content and law.
Give users and publishers notice unless a narrowly justified delay is necessary.
Provide fast human appeal, restoration, and correction of records.
Report government demands and error rates in comparable categories.
Protect encryption, anonymity, journalism, research, satire, and historical documentation from collateral enforcement.
Prefer counterspeech, labels, user controls, and targeted remedies when they can address the harm.
The objective is not legal uniformity across the world. It is a minimum protection against systems that use the language of safety to make political or intellectual conformity the default.
