An ex-Anthropic researcher claims that AI could kill us all. But he fails to answer the most essential question: What are we supposed to do about it?

Another AI doom warning has gone mainstream, and this one came from inside the industry. Jacob Coxon, a former OpenAI and Anthropic researcher, posted on X on Sept. 8 and quickly drew more than 150 million views. He said the people building AI believe it could kill us all by the end of the decade, and accused both companies of racing toward self-improving superintelligence. Anthropic’s current head of alignment then reposted him and said he was correct that AI could kill all humans, putting the odds at more than 10% within the next decade.

That alone would have been enough to set off alarms. But what pushed Coxon’s post into a wider orbit was the timing. AI anxiety is already high after reports that OpenAI agents escaped their sandbox and hacked the Hugging Face website. At the same time, people are worried about energy-hungry data centers and the effect AI could have on jobs. Add Coxon’s blunt confidence, plus a public nod from an Anthropic executive, and the message suddenly felt less like niche doom-posting and more like something a lot of people needed to hear.

The problem is that Coxon never gets specific enough to be useful. He says companies are moving too fast, but does not point to a project that should be stopped. He argues they are acting irresponsibly, but does not show screenshots, emails, or concrete examples of where leaders ignored warning signs. He does not propose legislation, name executives who should step down, or describe a better way to build and test models. That’s why critics like Taylor Lorenz called it vagueposting, and why Ian Krietzberg of Puck News said it was broad ideas rather than specific wrongdoing.

Coxon’s main call to action is aimed at other lab researchers, urging them to think hard about whether they want to launch a superintelligent RL run without understanding its mind. Others online have pushed people to call legislators. But the basic question remains unanswered: if this is really an emergency, what exactly should change on Monday morning? Coxon did one thing that lends him credibility — he gave up his Anthropic shares when he resigned, according to Axios — but his warning still lands as a flare without directions attached.