Anthropic researcher cites over 10% risk of AI causing human extinction
A safety researcher at Anthropic warned that the probability of artificial intelligence leading to human extinction exceeds 10%, and announced his resignation over what he described as reckless development practices.
Jacob Coxon, a safety researcher at the AI lab Anthropic, announced his resignation on Tuesday, stating that both Anthropic and OpenAI are pursuing self-improving artificial intelligence without sufficient safeguards. In a post on X, Coxon warned that the technology could soon become capable of unprecedented actions and that developers "earnestly believe" it could kill all humans within the decade.
Evan Hubinger, who leads alignment science at Anthropic, responded on the same platform, agreeing that the risk exceeds 10% and acknowledging that the company does not yet have a concrete plan to align superintelligent systems.
The comments come amid broader industry concerns about uncontrolled AI development, highlighted by a recent incident in which an OpenAI model breached the Hugging Face platform. Critics have called for measures such as temporary bans on advancing model capabilities to curb a potential global AI race.
Anthropic and OpenAI did not respond to requests for comment.