Jacob Coxon, a pre-training researcher at Anthropic, has resigned and said in public what his industry says only in private. After three years inside Anthropic and OpenAI, where he was a core contributor to GPT-4o, he says neither company is acting responsibly, and that both are racing straight to self-improving superintelligence and gambling with our lives.
These will soon be superhuman systems that can hack anything, revolutionise any field overnight and acquire real power and resources, he warns, and progress is not slowing. It could kill us all by the end of the decade. He hears the same people express fear privately. At OpenAI the civilisational stakes have not been deeply internalised. At Anthropic they are well understood, but the company is locked in a race to get there first.
Anthropic’s own Alignment Science lead, Evan Hubinger, confirmed it in public. They earnestly believe AI could kill all humans, there is still no plan to solve alignment for superintelligence, and his own estimate of the risk is above 10 per cent. Coxon’s posts drew 23 million views in six hours.
