Jacob Coxon, a 27-year-old British researcher who worked for three years at OpenAI and subsequently at Anthropic, decided to leave the artificial intelligence industry with a warning that set off alarm bells.
Shortly after resigning from Anthropic, Coxon posted on X that neither company is acting responsibly and that both are racing full speed ahead toward a self-improving superintelligence.
His concern is even more extreme: he claimed that those developing these technologies sincerely believe that an advanced AI could kill all humans before the end of this decade.
According to Coxon, these systems could far surpass human capabilities, hack computer systems, revolutionize entire fields, and gain access to resources and power in the real world.
But what makes his statements especially striking is that he says he has heard executives and researchers privately express fears far greater than those they voice publicly.
And there was a response that gave his words even more weight.
Evan Hubinger, an Anthropic researcher and Coxon’s former boss, publicly responded that “we really believe” that AI could kill all humans, although he estimated that he personally considers that risk to be below 10%. He also acknowledged that Anthropic still has no plan to solve the safety problem of a future superintelligence.
Coxon argues that the problem is the race between companies: even those who understand the risks feel they must move forward before their competitors.
And that leaves the question raised by his warning: if the very people building the technology believe it could become dangerous, are we moving too fast?