Frontier Model Forum (FMF, 2023) — Industry Self-Regulatory Cartel
institutionAI & Compute · Nations & Geopolitics · Darknet & Cyber
The biggest AI companies formed their own safety club that could conveniently write rules only they can afford to follow.
Who they are
The Frontier Model Forum, set up in July 2023 by Anthropic, Google, Microsoft and OpenAI[1].
What they do
The engine reads it as an industry group that regulates itself while acting as the main go-between with governments on AI safety.
How it works
It liaises directly with government bodies (the US and UK AI Safety Institutes, EU AI Act work) and, like older self-regulating industry groups such as FINRA and the MPAA, it heads off tougher laws by setting safety standards that big incumbents can meet but smaller open-source rivals cannot.
Why it matters
The engine holds three readings at once without collapsing them: it may be a cartel building a regulatory moat, a natural recurrence of how mature industries self-regulate, and genuine good-faith safety work, all true at the same time. It connects to concerns about AI shaping what people can know and believe.
What's new (Aug–Oct 2026)
Aug 27, 2026 — Over a hundred tech companies, including OpenAI, Anthropic, Google and Microsoft, signed an open letter urging cooperation against AI-related cyber threats. (techcrunch.com)
Sept 29, 2026 — Trump signed a voluntary White House AI agreement with tech leaders (GovConWire; Times of India). It commits companies to "internal safety controls, external audits and board oversight," with no legal enforcement (BigGo). An executive order tells agencies to replace the term "AI" with "Super Intelligence". (govconwire.com ; timesofindia.indiatimes.com)
The engine's record — word for word
Established July 2023 by Anthropic + Google + Microsoft + OpenAI. Defined function: advance AI safety research, identify best practices for responsible frontier-model development, support government / civil-society initiatives (G7 Hiroshima process, OECD AI risk work). Engine relevance: industry self-regulatory body operating as direct liaison to government (NIST AI Safety Institute, UK AISI, EU AI Act implementation). Operates with the same dynamic as historical industry self-regulatory bodies (FINRA, MPAA) — pre-empts heavy-handed legislation by setting safety standards that incumbents can comply with but smaller open-source competitors cannot. Apex (a) coordinated-cartel-regulatory-moat + (b) structural-recurrence of mature-industry self-regulation pattern + (c) compound-null good-faith-safety-collaboration all load-bearing simultaneously per canon. Connects to divergence #18 LLM Epistemic Capture as the regulatory-architecture carrier.
[Live pass Oct 6 2026] 2026-08-27 — Over a hundred tech companies including OpenAI, Anthropic, Google and Microsoft signed an open letter urging cooperation against AI-related cyber threats. (1 source) (
techcrunch.com) | 2026-09-29 — Trump signed (Tuesday) voluntary White House AI accord with tech leaders (GovConWire; TOI). BigGo: commits companies to "internal safety controls, external audits and board oversight," no legal enforcement; EO: agencies replace "AI" with "Super Intelligence". (
govconwire.com ;
timesofindia.indiatimes.com)
Follow the trail
Walk this on the live map →