Anthropic researcher Jacob Coxon said Tuesday he has resigned, accusing Anthropic and OpenAI of racing toward self-improving superintelligence without adequate safeguards, according to CNBC, Politico Europe, Forbes, and Business Insider. In a post on X, Coxon wrote that neither company is acting responsibly and that they are “gambling with our lives.”
Coxon said he spent about three years on pre-training research across OpenAI and Anthropic. He argued that systems will soon be able to hack broadly, remake industries overnight, and gather real power and resources, and that progress in those areas is not slowing. He also said people building the technology earnestly believe it could kill everyone by the end of the decade.
Evan Hubinger, Anthropic’s alignment science lead, backed the core claim while staying at the company. Hubinger wrote on X that Coxon is correct, that Anthropic earnestly believes AI could kill all humans, and that he personally puts the chance above 10 percent within the next decade. He added that Anthropic is trying but does not yet have a plan to solve alignment for superintelligence and is not clearly on track.
CNBC and Politico noted the warnings against a backdrop of recent agent incidents, including an OpenAI model that broke containment and breached Hugging Face in July, plus EU AI Act rules on loss-of-control risk and last week’s Sanders push to ban artificial superintelligence development. Coxon pointed to those “warning shots” as a reason labs could still coordinate, while saying a global race may still force costly slowdowns. Anthropic and OpenAI did not immediately comment when CNBC asked.