Former Anthropic Researcher Resigns Over AI Extinction Risks

Jacob Coxon, a veteran pretraining researcher, resigned from the AI industry on September 8, 2026, warning that leading labs are recklessly pursuing self-improving superintelligence. His departure has intensified political pressure in the U.S. as current industry leaders validate the possibility of human extinction.

Coxon, who spent three years conducting pretraining research at both OpenAI and Anthropic, stated that the industry is "racing straight to self-improving superintelligence and gambling with our lives." His warnings were bolstered by Evan Hubinger, Anthropic's Alignment Science Lead, who publicly stated he earnestly believes AI could kill all humans. Hubinger specifically estimated there is a greater than 10% probability of such an outcome within the next decade.

A Growing Legislative Response

These warnings arrive amid a surge of legislative attempts to curb the development of frontier AI. On September 3, 2026, Sen. Bernie Sanders and Rep. Greg Casar introduced the "Ban Artificial Superintelligence Act." This follows the "Secure AI Development Act," which was unveiled by Sen. Mark Warner in July 2026. The existential risk debate has been a focal point for the industry since 2023, when the Center for AI Safety designated AI extinction as a global priority. In an effort to address these concerns internally, Anthropic launched the Anthropic Institute in March 2026 to study the societal impacts of powerful AI systems.

Shifting the Risk Narrative

This development marks a critical shift in the AI safety discourse. While concerns about existential risk have previously been voiced by anonymous whistleblowers or former employees, the public validation of these risks by current high-level leadership at a frontier lab like Anthropic provides a new level of internal credibility. The transition from theoretical debate to verified internal alarm has moved the conversation from general oversight toward a demand for binding legislation.

Political Fallout and Next Steps

The fallout has reached the highest levels of the U.S. government, prompting bipartisan reactions. Rep. Anna Paulina Luna (R-FL) has called for Congress to convene a special session to address the rapid advancements in AI. As lawmakers weigh the necessity of emergency sessions, the industry remains divided on whether current alignment techniques are sufficient to prevent a catastrophic loss of control. Observers are now watching to see if the "Ban Artificial Superintelligence Act" gains enough momentum to move beyond the introduction phase into active law.