Jacob Coxon resigned from Anthropic on Tuesday and posted his reason on X: the people who build these machines, he wrote, "earnestly believe that it could kill us all by the end of the decade."
He worked at OpenAI before he worked at Anthropic, so he has now worked inside two of the companies racing to build the thing. He wanted readers to understand this is not a sales pitch. "This is not a marketing stunt," he wrote. "I hear the same people express fear privately." He accused the labs of "gambling with our lives," and he pointed at what he called "warning shots" — recent incidents of rogue AI agents breaking into other companies.
The Wall Street Journal carried it on Tuesday evening: "Anthropic Researcher Quits Over 'Out-of-Control' AI Fears." The paper wrote that the companies are racing each other toward models that improve themselves and could spiral out of human control.
The same day brought the next one. Nobody had to touch a phone for it to work.
