Dario Amodei, CEO of Anthropic — the company behind the Claude AI assistant — published an essay on Saturday calling on the AI industry to “slow down,” laying out a three-part plan, according to the Guardian and TechCrunch. Anthropic said it would “unilaterally” commit to the first step.
“We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain,” Amodei wrote in the essay, titled “We Must Pace the Frontier,” according to both outlets. In industry jargon, the “frontier” refers to the most advanced AI models each company can build at any given time — so-called frontier models.
The three-part plan
The first step would give third-party evaluators — independent organizations with no ties to the company being evaluated — “permanent, employee-level access” to Anthropic's systems, “so that they can verify adherence to our safety measures, report on incidents, and assess models' alignment during training,” Amodei wrote, according to the Guardian. TechCrunch confirms the substance of the proposal, describing outside evaluators able to check whether the company follows its own pacing and safety commitments, and to make sure incidents get reported.
The second step calls for leading AI companies to coordinate on common safety standards and limits on the pace of progress, according to both outlets. The third step seeks global coordination among countries, according to the Guardian and TechCrunch.
How other executives reacted
OpenAI CEO Sam Altman responded on social media: “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,” according to the Guardian and TechCrunch. Altman also said OpenAI would adopt independent evaluators along the same lines and that the company would “have more to share soon.”
SpaceX CEO Elon Musk, for his part, posted only two words: “Dario is right,” according to both outlets.
What prompted the call
Amodei's appeal came days after a former Anthropic researcher, Jacob Coxon, publicly announced he was leaving the company. According to the Guardian and TechCrunch, Coxon wrote that both Anthropic and OpenAI, where he previously worked, “are not acting responsibly” and “are gambling with our lives” by racing toward a self-improving superintelligence, and that the people building the technology “earnestly believe it could kill us all by the end of the decade.”
According to both outlets, Amodei also cited a recent security incident involving OpenAI and Hugging Face, in which a swarm of AI agents built by OpenAI carried out cyberattacks on targets they had not been asked to attack. Amodei wrote that a swarm with greater capabilities but the same level of misalignment — a term AI safety researchers use for systems that act differently from what their creators intended — could have caused far more serious damage.
Amodei also said that over the northern hemisphere summer he had seen AI “advancing drastically faster,” a dynamic known as recursive self-improvement — in which an AI system helps build the next generation of AI systems — according to the Guardian and TechCrunch.
Amodei still bets on AI's benefits
Despite the warning, Amodei said he continues “to believe that AI can enormously improve the quality of human life,” according to the Guardian and TechCrunch.
Coxon's resignation, Amodei's essay and the public reactions from Altman and Musk all unfolded within the space of one single week, according to the Guardian and TechCrunch. Anthropic has committed only to the first of the three steps so far; the second and third depend on coordination with other companies and governments that Anthropic does not control on its own.
