Content Moderation — The Control-Grid Component

Content moderation is the soft end of internet control: not blocking the network, but deciding which speech survives on it. The distinction that matters is who decides and under what review — a state pre-clearing posts, a regulator mandating removal timelines, or a private cartel of platforms and advertisers setting de-facto global rules with no appeal. Each jurisdiction sits at a different point, and the trend everywhere is toward faster removal with less process. Scored in the Convergence Index.

China

China pushes prior restraint onto the providers. The Cyberspace Administration of China combines licensing, real-name rules, and provider-side censorship duties so that much content is filtered or cleared before it circulates rather than removed after (CAC) — with no notice, no appeal, and no line between platform rules and Party rules. This is the maximal form: the moderation obligation runs ahead of the speech.

European Union

The EU mandates removal by statute. The Digital Services Act imposes systemic-risk and removal duties on large platforms (DSA, Regulation (EU) 2022/2065), and the Terrorist Content Regulation requires removal within one hour of a competent-authority order (Regulation (EU) 2021/784). These carry transparency obligations and judicial review — the bounded version — but a one-hour mandate with statutory fines pushes platforms toward over-removal as the safe default.

India

India pairs fast takedown with a check that occasionally bites. The IT Rules 2021 give the government rapid takedown powers over intermediaries (MeitY) — but when the government created a Fact Check Unit empowered to flag “fake” content about its own affairs for removal, the Bombay High Court struck it down as unconstitutional (LiveLaw). The takedown machinery is broad; the judicial friction is real but intermittent.

United States

The US has no state moderation mandate — and built a private one instead. The function migrated to a cross-platform layer: the Global Alliance for Responsible Media coordinated advertiser pressure on platform policy until it was discontinued on August 9, 2024 — three days after X filed an antitrust suit naming it (WFA / GARM); the Global Internet Forum to Counter Terrorism runs a shared hash-database that removes flagged content across member platforms at once (GIFCT); and the Christchurch Call coordinates governments and platforms on terrorist-and-violent-extremist content (Christchurch Call). The Twitter Files documented the informal government-to-platform flagging channel that operated alongside these. No First Amendment problem, because the state is not formally the actor — which is exactly what makes the private layer hard to appeal. The personnel and NGOs that staff this private layer — the Stanford Internet Observatory, the GEC, GDI, CCDH, the ADL, and the Trust & Safety profession itself — are documented in depth in the Fires series’ content-moderation infrastructure file.

United Kingdom

UK users fall under the Online Safety Act 2023’s statutory duties, enforced by Ofcom as the statutory regulator.

The counter-argument

Some moderation is non-negotiable: CSAM, real-time terrorist livestreams, and coordinated foreign influence operations are not “speech” disputes, and the shared-hash systems that kill a Christchurch-style livestream in minutes are a genuine good. The capture risk is scope creep — the same hash-sharing and one-hour-removal machinery built for the unarguable cases becomes available for the arguable ones, with the appeal step quietly missing. The component scores the machinery; the missing appeal is what the “limit” axis tracks.

Part of the Convergence Index component set — scored in the convergence table.

Related research

• AI moderation — the mechanism/bias deep-dive
• Deplatforming — escalation to stack-level removal
• The Twitter Files · Internet control