Three researchers associated with Anthropic went public overnight with stark warnings that rapidly advancing, potentially uncontrollable artificial intelligence (AI) could pose an existential risk to humanity within the coming decade.
What the researchers said
-
Jacob Coxon, who resigned on Tuesday, wrote on X (formerly Twitter): “The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible - but I hear the same people express fear privately. No other human activity poses this level of danger.”
-
Evan Hubinger, Anthropic’s alignment‑science lead, responded: “Jacob is correct here — we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. 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.”
-
Samuel Marks, Anthropic’s scalable‑oversight lead, added: “AI developers believe their technology could cause human extinction (or similarly bad outcomes). This could happen in the next few years. In general, the more senior the employee, the more concerned they are.”
Why it matters
These statements echo other recent warnings from inside the industry. Days earlier, leaders at OpenAI, including CEO Sam Altman, had said that AI is advancing into a frightening, hard‑to‑control phase. Internal warnings across major AI firms highlight a shared dilemma: slow down and risk falling behind competitors, or continue to push capabilities and risk losing control.
Reporters Maria Curi, Madison Mills and Ina Fried at Axios note that companies continue to develop models aggressively even as they call for regulation or global pauses — a stance critics argue may entrench dominant firms.
The broader picture
Calling the situation unprecedented is not an overstatement: some of the fastest‑growing technology companies in history are warning that their own products could cause severe harm or even human extinction. According to conversations with multiple insiders, many staff who have seen unreleased models are increasingly alarmed, and their concerns should not be dismissed out of hand.
Considerations for policymakers
The warnings are difficult to validate directly, but commentators (noted as Mike & Jim in the reporting) argue that lawmakers and relevant federal agencies should be aware that AI creators themselves see the potential for catastrophic outcomes unless the processes by which the United States, China and others review and release more powerful models change.
How to think about the risk
Researchers are not claiming AI is an imminent, certain global threat today. Their point is that AI is improving faster than expected and may soon enable recursive self‑improvement. If systems reach that stage, their capability growth could accelerate and become far harder to predict or control.
