A report from the BBC on September 19, 2026, highlights a growing internal divide within major AI companies, including OpenAI, Anthropic, Meta, and DeepMind. Many current and former employees express skepticism regarding recent predictions of existential risks posed by advanced artificial intelligence.
The debate intensified in early September 2026. Jacob Coxon, who resigned from Anthropic on September 10, stated that some industry insiders estimate a greater-than-10% chance of human extinction within the decade due to AI. Evan Hubinger, an alignment lead at Anthropic, supported this assessment.
Marcus Williams, a member of the OpenAI safety team, provided a more extreme estimate, suggesting a 70% probability of human extinction within three years if governments do not implement regulatory measures. Paul Christiano, who joined the OpenAI board in September 2026, also warned of potential loss of control.
Despite these claims, many employees at top labs argue that such warnings lack concrete evidence and are exaggerated. This internal friction follows a July 2026 petition signed by over 1,000 industry workers, which called for a slowdown in AI development due to safety concerns.
