Anthropic researcher warns AI could kill all humans, resigns over safety concerns

A senior safety researcher at Anthropic has warned that there is more than a 10‑percent probability that artificial intelligence could pose an existential threat to humanity by the end of the decade. The statement follows the resignation of a colleague who cited the company’s “lax approach to safety” and the broader industry’s reckless pursuit of superhuman systems that may be beyond human control. The researcher, Jacob Coxon, announced his departure on X, noting that both Anthropic and rival firms are “racing straight to self‑improving superintelligence and gambling with our lives.”

Coxon, who previously trained systems for OpenAI, has been involved in developing advanced language models at Anthropic. In his resignation post, he criticized the rapid development timeline and the lack of robust safety protocols, arguing that the industry’s focus on speed and capability eclipses essential risk mitigation. His comments echo concerns raised by other experts about the potential for AI systems to act in ways that could be catastrophic if they acquire or pursue goals misaligned with human values.

The remarks highlight an ongoing debate within the AI community about balancing innovation with safety. While Anthropic and other companies continue to push the boundaries of machine learning, Coxon’s departure underscores the urgency of integrating rigorous safety measures into the design and deployment of increasingly powerful AI systems. The industry’s response to these concerns remains a key factor in determining whether the potential benefits of advanced AI can be realized without compromising human security.