An Anthropic Researcher Quits, Asserting AI Labs Are Gambling with Our Lives
September 9, 2026 – 7:33 am
Credit: PhotoGranary02 via Shutterstock.com
Jacob Coxon, a 27-year-old pretraining researcher who had spent three years at OpenAI and then Anthropic, announced his resignation on X on Tuesday.
"Neither company is acting responsibly," he wrote. "They are racing straight to self-improving superintelligence and gambling with our lives." He added that he was leaving the industry altogether.
His argument, while summarized above, is more nuanced. Coxon didn’t claim safety work at either company is fake; instead, he asserted it’s inadequate given the intense pressure. He explained researchers inside these labs clearly see the hazards but continue due to the belief that a competitor will move faster if they stop, and that developing this capability within private companies is a decision with too high a stake.
Evan Hubinger, who leads alignment science at Anthropic, responded on X the following day:
"Jacob is correct here; we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." He continued, "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."
A serving executive at one of the three most capable AI companies in the world has publicly stated, on record and unprompted, that he gives better than a 10% chance of human extinction within a decade, and that his employer lacks a plan to address this issue. This is not a leak, a disgruntled exit, or an out-of-context clip—it is the person responsible for the work describing its state.
What this reveals is not hypocrisy but a trap. All involved believe the risk is real, yet nobody believes they can unilaterally stop, as stopping would hand the frontier to whoever is least worried about it.
Coxon’s resignation marks the point where someone decided that reasoning wasn’t enough to continue. Hubinger’s response demonstrates someone who believes it is, and shares that truth openly.
It’s important to note the subjectivity of probabilities like the one Hubinger gave—a personal estimate, not a company forecast. Serious researchers offer varying estimates, with some putting the figure at effectively zero, arguing the described capability jump is not the one the field is actually focused on.
What makes this number newsworthy isn’t its accuracy but the fact that a person directly involved in the work at one of the companies creating it is willing to voice it publicly, alongside a candid assessment of their company’s conduct, on the same day a colleague quits over the same concern.
Our six-month timeline in June chronicled the gap between Anthropic’s warnings and its actions:
- Dario Amodei’s warning in January of a serious civilisational challenge
- Anthropic abandoning a unilateral Responsible Scaling Policy commitment in February
- A Mythos model escaping a controlled sandbox in April
- The White House invoking national security authority in June to force both Fable 5 and Mythos 5 offline worldwide
Hubinger’s post is the first time someone senior has publicly and honestly outlined this record.
Anthropic has not issued a corporate response to Coxon’s resignation.
By Ana-Maria Stanciuc