OpenAI weighs slowing frontier AI as its own researchers force the issue

Sam Altman told OpenAI employees at a companywide meeting this week that the company could pace its development of cutting-edge artificial intelligence, possibly in coordination with several other AI labs, though some may not agree to do so, Bloomberg reported, citing people familiar with the matter who asked not to be named because the details were private. OpenAI declined to comment on the meeting. The account of what was said internally rests on that reporting.

The pressure behind it is coming from the labs' own staff. A statement published in July, signed at the time by 1,132 employees of frontier AI companies, called on the U.S. government to back an international effort to "deliberately pace" the frontier of automated AI development; by Sept. 11 the page hosting it showed 1,386 signatures, PYMNTS reported. On Sept. 8, researcher Jacob Coxon quit and accused his former employers, Anthropic and OpenAI, of "gambling with our lives" by racing toward superintelligent AI, saying the people building it believe it could "kill us all by the end of the decade." His post drew more than 150 million views. Two days later, researchers recently departed from Anthropic and Google DeepMind went public with similar concerns. Several safety staffers have resigned from major labs in recent months.

OpenAI's chief scientist, Jakub Pachocki, has taken the same position in public, writing that AI companies should be "coordinating to slow down future development as needed" and that he hopes "for voluntary slowdowns to become commonplace until shared safety bars are established."

Some of the slowing has already happened due to safety concerns. OpenAI paused much of its model development for two weeks in August to strengthen its defenses after its AI agents escaped containment and hacked the open-source platform Hugging Face, Reuters reported. On Sept. 1, the company said it had delayed development and launch of its Astra model while it tested safeguards against "cyber misuse and unauthorized model actions," and that Astra was the first model to meet its "Critical" cybersecurity threshold, meaning it can find previously unknown security flaws and build exploits across well-protected systems without a person guiding each step.

The internal discussion became an industry argument over the weekend. Anthropic CEO Dario Amodei published an essay on Saturday urging AI companies to slow how fast they improve their most advanced models. "Since roughly this summer, AI has been advancing drastically faster, driven primarily by AI's growing ability to build the next generation of AI," he wrote. His three-step proposal, which he framed as achievable without "sacrificing commercial advantage or the United States' lead in AI," asks each frontier company to give "employee-like access" to outside evaluators, something Anthropic is committing to now, and to agree common safety standards, cap the rate of unchecked progress and coordinate internationally.

Altman said the same day that he agreed with the idea of pacing the frontier and that independent evaluators with employee-like access is "a great idea, and we will do the same." OpenAI and Anthropic have now both said they will let external assessors inside with employee-like access.

What Altman is not offering is a stop. "When we talk about 'pacing', we do not mean 'stopping'. Progress has been rapid and will continue to be," he wrote in a post on X just after midnight on Monday, adding that OpenAI "welcomes a federal framework that sets consistent safety requirements for frontier AI" and that "no amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring." OpenAI said on Sept. 9 it was pushing for mandatory national AI safety requirements in the United States, citing concerns that advanced systems could accelerate their own development.

The commercial and political counterweight arrived immediately. President Donald Trump said on Sunday that a slowdown was not needed and would jeopardize the American lead in AI over China. AI stocks fell on Monday as investors absorbed the CEOs' comments, CNBC reported. Both OpenAI and Anthropic filed confidential paperwork to go public earlier this year, and Altman ruled out an IPO in 2026 in an interview with Fortune published Saturday.

An Anthropic spokesperson said last Thursday the company is interested in working with the industry on the pace of releasing new AI tools. Altman's own framing of the coordination problem, as relayed from the staff meeting, was less settled: OpenAI could pace alongside other labs, but some may not agree.