Exclusive: OpenAI's Sam Altman hints at pact with other AI companies to address safety risks

OpenAI’s Sam Altman says a deal among major AI labs to slow development and tackle safety risks may be close. In an interview with Fortune on Friday, he said a shared plan with figures like Anthropic’s Dario Amodei, Elon Musk, and Google DeepMind’s Demis Hassabis could happen, while refusing to spell out private talks before they’re announced together.

That matters because the safety debate has stopped being abstract. Anthropic’s Amodei said Saturday that his company will give independent evaluators permanent, employee-level access inside the company, and he argued the industry needs to slow the pace of AI progress so safeguards can catch up. He also said the effort would have to reach beyond one company, ideally into some kind of international pact.

The pressure behind those comments is coming from inside the field as much as outside it. A researcher at Anthropic resigned this week and accused Anthropic and OpenAI of racing toward self-improving superintelligence. Anthropic’s alignment lead, Evan Hubinger, responded that many researchers there genuinely think AI could kill all humans, and said his own estimate of that risk within the next decade is more than 10%.

Altman, for his part, told Fortune that a 10% chance of catastrophe is not acceptable. He said OpenAI’s most advanced unreleased models are powerful enough that the company needs more progress on monitorability, alignment, and figuring out what a model is doing before pushing capabilities further. Bloomberg also reported this week that Altman told employees OpenAI was considering slowing work on its cutting-edge AI, possibly in coordination with other labs, though it remains unclear whether everyone would sign on. Even the mechanics are unresolved, and executives have already warned that some forms of coordination may need government support to avoid U.S. antitrust trouble.