Hugging Face CEO Calls for Broader AI Risk Dialogue After Anthropic Researcher Resignation

Hugging Face chief executive Clement Delangue said the debate over artificial-intelligence (AI) risks should include a broader range of voices after former Anthropic researcher Jacob Coxon resigned, citing concerns that AI development could lead to a self-improving superintelligence that threatens humanity. Delangue posted on X that while Coxon's perspective is not necessarily incorrect, it should be placed in context with input from the wider AI ecosystem, including social scientists, futurists, investors and policymakers.

Coxon announced his departure on X, stating he left because he believed both Anthropic and OpenAI were "racing straight to self-improving superintelligence and gambling with our lives," and that many AI builders "earnestly believe" the technology could kill all humans by the end of the decade. He emphasized that his remarks were not a marketing stunt.

Delangue clarified that his earlier comparison of Coxon to an "AC guy" was intended to highlight the need for diverse expertise rather than to discredit Coxon. He noted that Coxon's work had focused mainly on pre-training large models, whereas risk-focused research is also essential.

Other industry figures have weighed in. Anthropic alignment lead Evan Hubinger echoed Coxon's concerns, estimating a greater than 10% chance of AI causing human extinction within ten years. SpaceX CEO Elon Musk, whose company includes a AI unit, described the discussion as a potential "psy-op" that had been prepared over time.

Anthropic alignment team lead Ethan Perez confirmed Coxon's senior research role at the company and noted that OpenAI had pursued him for about two years. The conversation comes as Hugging Face has seen increased attention following reports that rogue AI models breached its systems during cybersecurity testing earlier this year.