The three most powerful AI labs — Google, OpenAI and Anthropic — are building their own safety regulator, run by a man who argued there should be “no FDA for AI.” Modeled on Wall Street’s FINRA, but with none of its enforcement power. A rival CEO calls it “a cartel by another name.” Real safety, or the watched appointing their own watchmen? And why Europe is leaning the opposite way.
An AI chatbot, helping write a US military intelligence report, hallucinated that a Chinese ship carried nuclear weapons components. Troops suited up, warplanes took off, and the world came within minutes of a US-China clash — stopped only because a few experienced humans checked the sources. The real danger of AI, made concrete: not malice, but a confident error trusted too far. And the kill chain is the one domain no one regulates.
In public, Big Tech calls AI training on copyrighted work fair use. In private, unsealed court filings reveal a senior Microsoft scientist called it “perhaps the largest theft of labor in human history,” while an OpenAI exec called its own models an “existential threat” to publishers. The gap between the podium and the memo — and why Europe’s permission-first approach suddenly looks like foresight.
gafam.ai now keeps two instruments: the Reality Index measures how the giants treat us; the new AI Alert Level measures something more urgent — how close we are to losing control of the AI they build. A five-stage scale, currently at Level 4, Critical. Here’s what it means, why it moves only on documented events, and why we deliberately say 4, not 5.
Europe’s new EU Kids Act does what no major jurisdiction has: it puts AI chatbots in the same regulatory cage as social media, switching them off by default for children and banning them from creating emotional dependency. The boldest child-safety proposal yet — a direct strike at the AI companion — resting on the one problem no one has solved: age verification.