AI leaders call for restraint, but no company has delayed a frontier model

Several leading AI companies are backing a call to slow frontier development, but none has identified a model it will delay. Anthropic CEO Dario Amodei proposed “pacing the frontier” in an essay warning that AI capabilities are advancing faster than safeguards. OpenAI CEO Sam Altman and Google DeepMind CEO Demis Hassabis endorsed the idea.

The proposal focuses on independent evaluators, monitoring, shared safety standards and international cooperation. It does not define whether slowing development means additional safety testing or a longer gap between major releases. Without a public release schedule, the meaning of “pacing” remains uncertain.

U.S. Vice President JD Vance said it felt unusual that frontier AI companies were asking the government to regulate them, calling the situation potentially a “Trojan horse.” His comments highlight the tension between calls for restraint and continued competition. Anthropic released Claude Fable 5.1 and Mythos 5.1, while OpenAI debuted GPT-6 Astra, before Amodei’s latest appeal.

Amodei has cited risks including cyberattacks, bioterrorism, economic disruption and systems that humans may struggle to control. Surfshark engineer Karolis Kaciulis said extreme warnings can distract from current harms such as phishing, deepfake fraud, automated scams and intimidation. The article argues that any slowdown proposal will need independent evaluation and measurable commitments to have practical force.