Anthropic chief executive Dario Amodei called this week for the AI industry to deliberately slow the rate at which it improves the capabilities of its most advanced models, a proposal that drew rapid public endorsement from two of his biggest rivals, OpenAI's Sam Altman and xAI's Elon Musk.
In a policy essay published September 13, Amodei wrote that "we must slow the pace at which we improve the capabilities of AI models," arguing that regulation and legislation move far more slowly than the technology itself and risk being overtaken by dangerous capabilities before governments can respond. He proposed concrete steps, including giving independent, third-party evaluators employee-like access inside AI labs to verify safety practices, and establishing common safety standards that would apply across companies rather than each firm setting its own.
Altman responded within hours: "I agree with Dario that we need to pace the frontier." Musk was more terse, writing simply, "Dario is right." The rare public alignment among three chief executives who are otherwise fierce commercial rivals underscored how seriously the industry is treating recent internal warnings about the trajectory of AI capability gains.
A Resignation That Amplified the Debate
The slowdown proposal followed the September 8 resignation of Jacob Coxon, a researcher who had spent roughly three years working on pretraining at both OpenAI and Anthropic. In his resignation post, Coxon wrote that both companies "are racing straight to self-improving superintelligence and gambling with our lives," a message that drew more than 90 million views within 24 hours. Evan Hubinger, Anthropic's head of alignment stress testing, publicly backed the substance of the warning:
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is more than 10% within the next decade.
Evan Hubinger, Anthropic
Not everyone in Washington shares that alarm. President Trump dismissed the slowdown calls on September 14, saying "very negative forces" were stoking fears about AI "that won't happen," and argued that maintaining American technological leadership should take priority, adding that "whoever wins AI wins." The remarks set the administration's posture squarely against the pacing proposal championed by three of the industry's most prominent executives.
Markets React, IPO Timelines Slip
Investors treated the episode as a genuine signal rather than posturing. AI-related stocks fell in the days following Amodei's essay while cybersecurity shares rallied, as traders weighed the prospect of slower capability growth against continued demand for defensive AI tooling. The debate has also bled into corporate timelines: Altman told Fortune that OpenAI's own listing plans have slipped in part because of the safety questions raised in this same debate, while Anthropic itself is separately preparing for a stock market listing even as its chief executive argues publicly for slowing down the technology it sells.
Amodei's essay, published on his personal site, goes further than the pacing request, proposing staged regulatory frameworks covering economic displacement, scientific-acceleration policy and controls on autonomous weapons, but it is the narrower call to slow capability gains that has resonated most with rival executives and markets alike. Whether the three companies translate the public statements into binding commitments, such as shared third-party auditing arrangements, remains unresolved, and no formal agreement between the labs has been announced.