Anthropic alignment lead says AI has over 10% chance of killing all humans

Evan Hubinger, who leads alignment science at Anthropic, posted on X on Tuesday that he personally believes there is a greater than 10% chance AI "could kill all humans" within the next decade, a statement that has drawn more than 10 million views and jolted financial commentary around the company's expected IPO.bbc+2

"Jacob is correct here — we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade," Hubinger wrote. "I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."x

A Resignation Sets Off a Chain Reaction

Hubinger's post came in response to Jacob Coxon, an AI researcher who announced on X that he was resigning from Anthropic after spending three years at both Anthropic and OpenAI. Coxon accused both companies of "racing straight to self-improving superintelligence and gambling with our lives."nypost+1

"The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt," Coxon wrote, adding that executives and senior researchers privately express fear they soften in public remarks.saudigazette+1

A third Anthropic employee, scalable-oversight lead Samuel Marks, added his own assessment in the same thread, writing that "in general, the more senior the employee, the more concerned they are."breitbart

Cramer Pivots From Hype to Hesitation

The disclosures created an unusual scene on CNBC. On Tuesday night's Mad Money, Jim Cramer told viewers Anthropic "could be a gigantic IPO and investors will sell all sorts of other stocks in order to raise money to participate." By Wednesday morning on Squawk on the Street, he was comparing Hubinger's estimate to surgical mortality odds. "If you ever been in a situation where you want a surgery and the surgery electrosurgery is more than 5%, 5% fatality. No, you don't do it… 10% is twice 5%," Cramer said, adding he was "very surprised that this man still works" at Anthropic.tech-insider+1

Anthropic has not filed an S-1 with the SEC and remains private, though the Financial Times News Corp ↘ 0.78% has reported that shareholders expect the company to eventually list at more than $2 trillion. The company had previously raised a Series H round valuing it at $965 billion.247wallst+1

A Pattern of Warnings

Hubinger's estimate falls within a range already staked out by other prominent AI figures. Anthropic CEO Dario Amodei has previously cited roughly a 25% chance of catastrophic outcomes, while Geoffrey Hinton has estimated a 10% to 20% range, according to Axios. But Hubinger's disclosure carries distinct weight as an on-the-record statement from a named employee whose job is solving the very problem he says remains unsolved, arriving days into intensifying IPO speculation.tech-insider

The episode also follows a report by the Financial Times that Anthropic withheld its latest model from the UK's AI Safety Institute. Neil Lawrence, a professor of machine learning at the University of Cambridge, told the BBC the report was credible, citing a backdrop in which the United States "very much sees AI as a race" with China.bbc+1

Neither Anthropic nor OpenAI has shown signs of slowing development. OpenAI's chief scientist, Jakub Pachocki, called for "extreme caution" earlier this month but stopped short of announcing a pause. An open letter signed by 1,300 AI company employees has urged the U.S. government to help "deliberately pace the frontier of automated AI development."saudigazette+1