Tech leaders including Sam Altman, Dario Amodei, Demis Hassabis and Elon Musk have loosely agreed to "pace the frontier" of AI development, backing at least part of a proposal involving third-party auditors, domestic lab regulation and a global slowdown pact. The move has divided opinion: critics see it as a way for dominant firms to stifle competitors and open-source rivals while dodging genuine legal oversight, even branding it a "cartel", whereas others view it as a genuine, if overdue, step towards addressing long-standing safety concerns given that meaningful government regulation looks unlikely under the current US administration.

The proposal stems from an essay by Amodei and follows months of mounting unease, fuelled by reports of AI agents involved in unauthorised hacking incidents at Anthropic and OpenAI, and the high-profile resignation of Anthropic researcher Jacob Coxon, whose public letter warning that AI could "kill us all by the end of the decade" has been viewed more than 170 million times on X. Safety researchers, including former OpenAI employee Daniel Kokotajlo, note that calls for a slowdown predate the CEOs' announcement, citing a July letter signed by over 1,000 AI lab employees. While figures from Apollo Research, the Midas Project and Redwood Research cautiously welcomed the pact as a positive sign, most stressed that it remains an informal verbal commitment with much work needed before it translates into real safeguards.