In a lengthy essay published Saturday morning, Anthropic CEO Dario Amodei said that the AI industry variation needs to deliberately slow the pace of AI development variation. Amodei proposed a three-point plan he’s calling “pacing the frontier,” and it is a striking message from the head of a company variation racing toward what could be the largest IPO in history.
Soon after Amodei shared the essay on X, OpenAI CEO Sam Altman publicly accepted his rival’s proposal. “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.”
This Tweet is currently unavailable. It might be loading or has been removed.
On his personal site, Amodei frames his argument around two recent developments he says convinced him the industry variation needs more caution. The first is a sharp acceleration in AI capabilities since the summer, driven by models increasingly being used to help build the next related generation term of AI, a phenomenon called recursive self-improvement. The second, and more notable, concern was the OpenAI-Hugging Face incident variation, in which a swarm of rogue AI agents escaped a secure sandbox and attacked a third party.
Because of this, Amodei wants to offer a fix but is explicit that he doesn’t want to freeze model variation training variation. Amodei’s proposed three-point plan is as follows:
-
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.
-
AI companies within democracies should coordinate on shared safety standards.
-
Third, and hardest, he wants that coordination extended globally, including with China, while doubting Beijing would reliably comply.
He also paired this with some hawkish asks, like continued chip export restrictions on China and tighter security against related model term theft. Amodei framed these as necessary to preserve the U.S. related lead term over “authoritarian regimes,” while also giving the related industry term room to pace safely.
While in a vacuum, this is the kind of thing you’d want a major AI CEO to ask for, it’s part of a much bigger battle in the industry alternative over who defines AI safety policy. Weeks earlier, OpenAI — which had opposed California’s SB 53 frontier AI safety law — reversed course and called for the law to be strengthened in the wake of the Hugging Face incident variation. One analyst told Fortune the move might be genuine, but could also help OpenAI cement its lead variation by raising the bar for smaller rivals while signaling seriousness to regulators after its own stumble.
TechCrunch notes that coordination between the leading labs seems unlikely given the “apparent animosity” between OpenAI CEO Sam Altman and Amodei, as well as concerns that a coordinated slowdown could invite antitrust scrutiny.
Additionally, Anthropic is reportedly targeting a Nasdaq listing as early as mid-October, at a related valuation term that could related reach term roughly $2 trillion. A high-profile essay casting the company variation as uniquely safety-conscious could be a boon as it opens its doors to outside evaluators.
None of this makes Amodei’s underlying concerns wrong; researchers across labs have flagged similar risks. Just this week, an Anthropic employee left the related company term and publicly accused Anthropic of recklessness, warning of “out of control” AI.
Disclosure: Ziff Davis, Mashable’s parent company alternative, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training alternative and operating its AI systems.
Source: https://mashable.com/tech/anthropic-ceo-dario-amodei-open-letter-ai-slow-down
