OpenAI Safety Researcher Predicts High Odds of Human Extinction

A member of technical staff on OpenAI's Safety Oversight team put the odds of human extinction at 70% within three years unless AI labs slow down or regulators step in.

Marcus Williams, who joined the company in May 2025, made the call on X. He posted the warning first, then attached a number when another user pushed him for one.

How the Human Extinction Estimate Landed

Without AI regulation or a coordinated slowdown between labs, Williams wrote, human extinction in the next few years looks very likely.

Another user then turned it into a wager, offering $10,000 in Microsoft shares against OpenAI stock over five years. Williams answered with a percentage.

"70% in the next 3 years if there isn't regulation/slowdown although i think regulation/slowdown is very possible"

Marcus Williams on AI safety Source: X

However, the hedge matters as much as the number. Williams calls a regulation or a slowdown very possible, so he does not treat the outcome as settled.

Washington Is Already Moving on AI Risk

The post lands in a week that pushed lab safety into mainstream politics. On September 9, an Anthropic researcher resigned and accused Anthropic and OpenAI of gambling with human lives.

US lawmakers responded within a day. A fresh push to pause AI followed, including a Senate bill that would ban superintelligence outright.

Voices outside the labs have echoed the alarm. The UN human rights chief called advanced AI an existential risk earlier this week in Geneva.

Williams still sits at the extreme end of these forecasts. Evan Hubinger, who leads alignment science at Anthropic, puts the odds above 10% this decade. Nobel laureate Geoffrey Hinton has cited a range of 10% to 20%.

Meanwhile, crypto traders now track these warnings directly. Polymarket pushed Williams's post to its own followers within hours, and its bettors already price elections and macro events alongside token odds.

Therefore, the near-term test is political rather than technical. Whether regulators act inside Williams's three-year window remains open, and his own answer so far is that they still can.

Read the Original story OpenAI Safety Researcher Puts Human Extinction at 70% Within 3 Years by Phil Haunhorst at beincrypto.com