AI Extinction Warnings Ignite Debate: Jacob Coxon's Resignation and the Push for Regulation

In early September 2026, former Anthropic researcher Jacob Coxon resigned from the AI company, citing fears that leading labs like Anthropic and OpenAI are racing toward self-improving superintelligence without adequate safety measures, potentially gambling with humanity's future. His X thread, which quickly amassed over 100 million views, claimed that insiders privately believe advanced AI could kill everyone by the end of the decade—a sentiment echoed by Anthropic's own alignment science lead Evan Hubinger, who publicly estimated a greater than 10% chance of human extinction from AI within the next decade.

Coxon's departure, first reported by the Wall Street Journal, triggered widespread media coverage and calls for regulation, including references to an AI non-proliferation treaty. Geoffrey Hinton, often called the 'godfather of AI,' told the BBC that a 10% risk estimate 'doesn't seem unreasonable.' The episode amplified existing discussions on existential risk, with outlets like WIRED and AP News highlighting internal industry concerns about 'crunch time' and the challenges of controlling rapidly advancing systems.

Skeptics, including figures like Elon Musk, questioned the timing and framing, suggesting elements of narrative coordination or a 'psy-op' to accelerate regulation amid competitive pressures. Coverage in Axios noted Coxon forfeited unvested equity by leaving early, underscoring the personal stakes. Broader context reveals this fits into ongoing debates: AI labs have long warned of risks, yet development continues at pace, raising questions about governance, international competition (especially with China), and public perception of technological existential threats.

The episode underscores how insider warnings can rapidly shape discourse on emerging technologies, blending documented industry anxieties with polarized interpretations of motives and timelines. While the 10% figure remains a personal estimate rather than consensus, it has intensified scrutiny on AI safety protocols and regulatory responses.