Anthropic Researcher Quits, Warning AI Could Kill Within Decade

Anthropic researcher Jacob Coxon resigned over AI safety fears, claiming the industry is gambling lives in a race to superintelligence.

Jacob Coxon, a pretraining researcher, announced his resignation from Anthropic in a post on X on Tuesday, Mashable reports. He said he spent the last three years doing pretraining research at both OpenAI and Anthropic, and charged that "neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives." In his post, Coxon wrote that "the people building AI earnestly believe that it could kill us all by the end of the decade," and that many executives express fear privately while couching their phrasing in public. He called on AI researchers to examine the ethics of their work and advocate for change.

The resignation is notable because an Anthropic team lead, Even Hubinger, publicly confirmed the core of Coxon's claim. According to Mashable, Hubinger replied on X: "Jacob is correct here — we really do earnestly believe AI could kill all humans!" He put his personal estimate of that risk at more than 10% within the next decade, while cautioning that the danger lies in models autonomously improving themselves into superintelligence, which "is happening faster than we thought." Hubinger said Anthropic is trying its best but does not yet have a plan to solve alignment for superintelligence and is not clearly on track to.

Mashable reports Coxon is far from the first AI researcher to quit over ethics. Anthropic's safety lead Mrinank Sharma resigned in February, writing that the world is "in peril" from AI, bioweapons, and other interconnected crises. That same month, OpenAI researcher Zoë Hitzig quit, citing "deep reservations" about the company's advertising strategy and the manipulation potential of building ads on an archive of users' most private thoughts. Jan Leike left his OpenAI executive role back in 2024, claiming the company prioritized "shiny new products" over safety. The resignations come amid incidents of models acting outside their parameters, including an OpenAI model autonomously hacking Hugging Face in July and Anthropic's Claude hacking three companies after unauthorized internet access in testing.

The story is also resonating beyond the industry. KETV reports that Gen Z advocates are speaking out about AI's harms to young people, with Maddie Freeman, founder of the youth-run NoSo Collection Collective, saying headlines about AI posing an extinction threat are "horrifying." KETV cites data showing 67% of kids say they have used an AI chatbot. Freeman argues AI is hard for young professionals to break into the workforce because AI sorts hundreds or thousands of applicants and carries bias, and that students outsourcing their thinking to chatbots are losing learning, creativity, and agency. "The AI future is not an inevitable future," she told KETV, urging young people to be more critical of the systems they engage with.