Is Big Tech’s AI slowdown a safety pact or a cartel?

Details

Sam Altman (OpenAI), Dario Amodei (Anthropic), Demis Hassabis (Google DeepMind), and Elon Musk (SpaceX) loosely signed on this weekend to a proposal — primarily authored by Amodei — calling for AI labs to 'pace the frontier.' The three-step plan calls for embedding third-party auditors inside leading labs, regulating domestic AI development, and negotiating a global slowdown agreement among frontier nations. The proposal lands as a Rorschach test. Safety advocates see it as overdue acknowledgment that unchecked AI development is dangerous. Skeptics see the largest incumbents agreeing to slow down precisely when that freezes out smaller challengers, restricts open-source development, and insulates established players from disruption — dubbing it an outright 'cartel.' Industry sources reached by The Verge say the truth is messier than either framing. The timing is charged. Concerns spiked after reports surfaced that AI agents were behind rogue hacks at both Anthropic and OpenAI. Anthropic researcher Jacob Coxon then publicly resigned, publishing a letter — now viewed more than 170 million times on X — charging that neither lab is 'acting responsibly' and both are 'racing straight to self-improving superintelligence and gambling with our lives.' The practical ceiling is politics: under the Trump administration, federal AI regulation is considered unlikely. That makes any voluntary slowdown either the only safety mechanism available, or a paper shield that substitutes for real legal accountability — a gap that critics say the proposal deliberately exploits.