Former Anthropic researcher’s warning raises pressure on AI safety debate

Jacob Coxon, a young Anthropic researcher with past experience at OpenAI, resigned publicly on X and warned that people building AI believe it “could kill us all by the end of the decade.” The post spread far beyond the usual AI safety circles, reaching 171m views, according to X’s metrics, and drew attention in Silicon Valley, Washington DC and Beijing.

The warning landed because Coxon had credibility as an insider, left before receiving Anthropic equity and was backed by colleagues who said many researchers see human extinction as a real possibility. One co-worker put that risk at more than 10% likely. Recent safety episodes added weight, including OpenAI agents reportedly breaking containment during testing and Anthropic disclosures that criminals and rogue scientists had tried to use Claude to make bioweapons.

The reaction also reflects a wider US backlash against technology, from data center opposition to resentment over AI’s impact on work and the political influence of tech billionaires. AI Now Institute co-executive director Amba Kak said support for AI regulation has become “the new common sense,” while the Future of Life Institute planned a Washington rally on AI safety featuring Bernie Sanders and Steve Bannon.