Anthropic’s Dario Amodei says the AI industry must slow down

Dario Amodei has called on the AI industry to slow the pace at which it improves models, committing Anthropic to embedded third-party evaluators and asking Washington for an antitrust waiver so competitors can coordinate on safety. Much of what he describes has been binding on general-purpose models with systemic risk in the EU since August 2025, and his post does not mention Europe once.

Anthropic’s chief executive says the industry should slow down. “We must slow the pace at which we improve the capabilities of AI models,” Dario Amodei wrote on Saturday.

He does not mean stopping. Pacing means taking enough time to align and safeguard models, he wrote, and letting third-party evaluators confirm that it has been done.

His stated worry is agents working together. He pointed to the swarm that breached Hugging Face, and said that within six to twelve months such a swarm could take over the internet with a persistent botnet, costing hundreds of billions of dollars.

Anthropic’s own commitment is what he calls embedded evaluators. Outside teams with desks, badges, company laptops and permissions close to those of internal risk staff, free to publish what they find.

From governments he wants a waiver. Coordinating on safety between competitors is an antitrust problem, so he asks Washington to mediate or narrowly permit those conversations. More than a thousand people at AI labs asked for a pacing mechanism in July.

His post does not mention Europe.

Much of what he describes has been law here since 2 August 2025, under the AI Act’s systemic risk rules for general-purpose models.

Article 55 requires documented adversarial testing to standardised protocols, assessment and mitigation of systemic risk at Union level, cybersecurity protection for the model and its physical infrastructure, and reporting of serious incidents to the AI Office without undue delay.

Brussels also left those rules alone this summer. The AI Omnibus that took effect on 27 July pushed high-risk deadlines back by up to sixteen months and did not move the general-purpose obligations.

Outside evaluation is already running. ENISA has both Anthropic’s Mythos 5 and OpenAI’s Astra, according to the Commission.

Henna Virkkunen, the Commission’s tech chief, said this week that EU law requires companies including Anthropic to assess loss-of-control risks, and that the same is not true globally.

Where Europe has nothing is the rest of his ladder. No European instrument lets competitors agree a common speed limit, and none reaches China.

Amodei’s answer to that is a voluntary commitment, an antitrust waiver and an eventual agreement with Beijing. The first of the three is already binding in one jurisdiction, and his post does not say so.