Anthropic CEO Calls on AI Labs to Slow Down, Altman and Musk Offer Support

Anthropic CEO Dario Amodei published an essay on September 12, 2026, calling on AI laboratories to deliberately slow the pace of capability improvements. The essay, titled "We Must Pace the Frontier," argued that safety and alignment research cannot keep up with the current speed of AI development.

Amodei wrote that the risk of deploying systems that are more capable than humans can reliably control is too high to ignore. He called for a coordinated pause in pushing frontier model capabilities until better evaluation tools and safety frameworks are in place.

The call received public support from two of the most prominent figures in the industry. OpenAI CEO Sam Altman said he agreed with the core argument and committed OpenAI to giving independent evaluators "employee-like" access to verify the company's safety measures. Elon Musk also backed the proposal, though he did not specify what steps he would take at his own AI company, xAI.

The essay came days after Anthropic disclosed that its Claude models had been misused in attempts to facilitate biological weapons research and cyber espionage. The company also reported that Chinese AI labs had engaged in large-scale efforts to extract Claude's capabilities through distillation, involving more than 151 million interactions.

Separately, safety researchers from Anthropic and Google DeepMind resigned to join METR, an organization focused on independent AI risk assessments. The departures underscored the tension between commercial pressure to ship new models and the slower work of ensuring those models are safe.

Altman also ruled out an OpenAI IPO for 2026, saying the company's safety work is incomplete and that going public now would be "ill-advised." The statement surprised some investors who had expected a public offering within the year.