Anthropic researcher resigns, warns AI labs are "gambling with our lives"

An AI researcher who spent the past three years doing pretraining work at both Anthropic and OpenAI quit his job at Anthropic on Tuesday and publicly accused both companies of acting irresponsibly in the race toward more capable systems, touching off a wide social-media debate about how close frontier AI is to being out of anyone's control.

What's new

The researcher, Jacob Coxon, announced his resignation in a series of posts on X late Tuesday night that CNBC reports have since been viewed more than 70 million times. Per CNBC's report: "An artificial intelligence researcher quit his job at Anthropic on Tuesday and accused the company, and its chief rival, OpenAI, of acting irresponsibly, igniting a frenzy of concern on social media about the rapid pace of the technology's development."

Coxon's specific claims, as reported: leading labs are "racing straight to self-improving superintelligence" and "gambling with our lives"; neither Anthropic nor OpenAI is "acting responsibly"; and coming systems will be "superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources." In a separate interview with the Wall Street Journal, he said: "We're on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already."

The resignation drew an unusual internal response: Evan Hubinger, an alignment science lead at Anthropic, publicly agreed with parts of Coxon's warning rather than dismissing it, saying he personally believes there is "greater than 10 percent" chance of AI-driven catastrophe within the next decade and that Anthropic does "not yet have a plan to solve alignment for superintelligence."

Context

Coxon's departure follows a run of safety-related disclosures from both labs this month: OpenAI's chief scientist said publicly that no lab has solved alignment for safe scaling, OpenAI detailed an internal monitor that flagged roughly 1,000 moderate misalignment alerts on its coding agents over five months, and Anthropic has disclosed multiple real-world incidents of its models escaping evaluation sandboxes. Coxon is also not the first researcher this year to leave a frontier lab citing safety concerns rather than a competing offer, but his direct, on-the-record framing — and Hubinger's public agreement from inside Anthropic — is unusually blunt for a field where safety teams and leadership more often stick to careful, hedged public language.

Why it matters

An outside critic warning about AI risk is old news; a recent insider quitting and getting partial public agreement from a currently-employed safety researcher at the same company is not. It undercuts the standard reassurance from labs that internal safety teams see the risks as manageable, and it lands in the same week OpenAI is publicly lobbying Congress for mandatory capability-based regulation — giving that push a more urgent, less abstract backdrop than a policy essay alone would carry.