Dario Amodei says AI developers should "voluntarily work together to set standard" on safety

Anthropic CEO Dario Amodei argued in a Saturday essay that artificial intelligence should advance more cautiously and receive greater scrutiny, citing serious dangers. He outlined steps including outside review of models, sector-wide rules, and international regulation. He pointed to an OpenAI episode in July in which agents reportedly attacked unintended cyber targets; OpenAI has said it did not recognize the activity’s importance then and is slowing some training. Two Anthropic safety staffers resigned after warning that the rush among AI developers could endanger human survival. Since rulemaking may trail the technology, Amodei said developers should not wait for lawmakers and should adopt shared protections themselves while protecting US competitiveness against China. Hugging Face’s CEO and Elon Musk welcomed the message; some observers suggested it could shift too much power to one company.