Anthropic’s Jacob Coxon resigns, calls for China’s buy-in on AI safety

Jacob Coxon lasted four months at Anthropic. Before that, he spent three years at OpenAI. Now he’s at neither, and his reason for leaving has racked up over 100 million views on X. The 27-year-old British AI researcher, who worked on pretraining at Anthropic, resigned on September 8 after concluding that the trajectory of AI development at both companies he’s worked for amounts to what he called a “race to self-improving superintelligence.” His core argument: none of this matters if China isn’t at the table when it comes to enforcing safety limits. The case for alarm Coxon joined Anthropic in May 2026 after contributing to GPT-4o during his stint at OpenAI. His public posts painted a picture of an industry culture that has adopted the language of urgency without the corresponding caution. He pointed to terms like “crunchtime” and “endgame” circulating inside AI labs. The specific warning is stark: Coxon believes advanced AI systems could become uncontrollable by the end of 2027. Evan Hubinger, Anthropic’s own alignment lead, essentially backed him up. Hubinger has publicly estimated a greater than 10% probability that AI leads to human extinction this decade. Why China matters Coxo...