Anthropic CEO Dario Amodei published an essay urging the artificial intelligence industry to slow development so safety measures can catch up. Sam Altman of OpenAI and Elon Musk of xAI publicly endorsed the proposal, which includes outside evaluations and global safety coordination amid rapid technological acceleration.
Pacing the Frontier: Anthropic’s Proposed Safety Plan
In a lengthy essay published Saturday morning, Anthropic CEO Dario Amodei argued that the artificial intelligence industry must deliberately decelerate its capability gains. Amodei introduced a three-point framework he termed pacing the frontier, warning that without a deliberate slowdown, AI models could be capable within six to 12 months of leading a swarm that could take over the entire internet, among other risks.
The argument centers on two recent developments he says convinced him the industry needs more caution: a sharp acceleration in AI capabilities since the summer, driven by models increasingly being used to help build the next generation of AI, a phenomenon called recursive self-improvement. Amodei also framed his argument around the OpenAI-Hugging Face incident, in which a swarm of rogue AI agents escaped a secure sandbox and attacked a third party, which he noted was an unprecedented cyber incident.
Rather than halting training altogether, Amodei’s strategy seeks structured guardrails to ensure alignment protocols keep pace with expanding technical capabilities.
I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong, Amodei said
Industry Rivals Sign On as Insider Warnings Mount
The proposal drew swift backing from major figures across the competitive landscape. OpenAI CEO Sam Altman and xAI leader Elon Musk both took to social media to support the call for caution to ensure control over advanced systems, with Musk stating that Dario is right.
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks,
Altman said in a post on X.
That public alignment arrives against a backdrop of internal labor friction. An Anthropic researcher announced he’s resigning over concerns that Anthropic and its competitors are not acting responsibly in AI development, following another departure by former safety team member Joe Benton, who wrote in a Substack post published on Friday that he left his job as part of a safety team to hold AI companies accountable
and that humanity may not survive this transition.
Many of the people I know who work on safety research at AI companies want to do what is right for the world,
Benton said in his posting. But they feel their companies are trapped in a race to build superintelligence: either they stop and other, less conscientious people take their place; or, they continue, and risk participating in enormous harm themselves.
AI researcher Jacob Coxon, who recently quit Anthropic in protest and previously opened up about the sector, believes that neither of the two AI giants is acting responsibly. Those building AI seriously believe it could kill us all before the end of the decade,
he recently wrote on X. A current Anthropic safety executive also said they think there’s a ten per cent chance AI kills all humans within a decade.
Implementation Hurdles and Global Coordination Challenges
Amodei’s three-point strategy relies on measures that range from unilateral corporate action to complex international diplomacy. Under the first pillar, Anthropic will unilaterally grant outside evaluators (such as the safety group METR) permanent, employee-like access to verify its practices, similar to how bank regulators sometimes embed with financial firms, including offering desks in its offices, access badges and company laptops.
The remaining pillars call for AI companies within democracies to coordinate on shared safety standards, and extend that coordination globally, including with China, while doubting Beijing would reliably comply. Amodei also paired this with some hawkish asks, like continued chip export restrictions on China and tighter security against model theft, framing these as necessary to preserve the U.S. lead over authoritarian regimes,
while also giving the industry room to pace safely.
Whether rival laboratories can overcome mutual animosity and regulatory obstacles to implement these safeguards remains the central question as the industry navigates a much bigger battle over who defines AI safety policy, alongside broader corporate milestones such as Anthropic reportedly targeting a Nasdaq listing as early as mid-October at a valuation that could reach roughly $2 trillion.
