Anthropic researcher quits, warns AI ‘could kill us all’

It’s a wonder that you got any work done after an artificial intelligence researcher who resigned from Anthropic this week accused both the Claude maker and OpenAI of acting irresponsibly and “gambling with our lives.”

“The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt,” the departed employee, Jacob Coxon, wrote in a post on X with 110+ million views. Coxon said that:

• Many people at OpenAI—where he worked before Anthropic—“have not deeply internalized the civilizational stakes,” but they’re “well-understood” at Anthropic.
• Still, Anthropic is “locked in a race,” and believes it must win because “no one else will act responsibly.” (Both companies are eyeing IPOs.)

Senior Anthropic leads might publicly pretend they don’t share these fears, Coxon said, but two of them essentially corroborated his statements on X. “We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,” one replied.

Déjà vu: Senior safety researchers at Anthropic, OpenAI, and xAI resigned at the beginning of the year, citing ethical concerns. Last month, Bill Gates issued a stark warning. In the past week, Sam Altman admitted conflicting feelings after the Hugging Face hack, and OpenAI’s chief scientist called for a slowdown of AI development—which some members of Congress also want to enforce.—ML