Marcus Williams, a technical staff member on OpenAI’s Safety Oversight team, stated that the probability of human extinction within three years is 70%. He noted that this outcome is likely unless AI laboratories implement a coordinated slowdown or regulators intervene. Williams joined the company in May 2025.
The estimate emerged following an exchange on X, where a user proposed a wager involving $10 000 in Microsoft shares against OpenAI stock over a five-year period. Williams clarified that he considers regulation or a slowdown to be very possible, meaning he does not view the extinction scenario as a settled outcome.
The warning follows a resignation on September 9 from an Anthropic researcher, who accused both Anthropic and OpenAI of risking human lives. In response, US lawmakers introduced a Senate bill aimed at banning superintelligence. Additionally, the UN human rights chief recently identified advanced AI as an existential risk.
Williams’s forecast is significantly higher than other industry estimates. Evan Hubinger, who leads alignment science at Anthropic, places the odds above 10% for this decade, while Nobel laureate Geoffrey Hinton has cited a range of 10% to 20%.
Market participants are increasingly monitoring these safety warnings. Polymarket, a prediction platform, shared Williams’s post with its users, as traders continue to incorporate such forecasts into their assessments alongside traditional macro events and token-based odds.
