Anthropic employees warn of greater-than-10% chance AI causes human extinction

Inside the labs building the most powerful AI systems in history, not everyone is losing sleep over the apocalypse. While a handful of prominent researchers have staked out dramatic positions on the existential risks of advanced AI, a broad swath of current and former employees at OpenAI, Anthropic, Meta, and DeepMind are greeting those warnings with something closer to eye-rolls than emergency protocols.

The BBC reported on September 19, 2026, that many AI lab workers reacted to existential risk warnings with skepticism and, in some cases, outright amusement.

The doomsday claims that started the argument

The tension crystallized around a series of high-profile warnings in early September 2026. Jacob Coxon, a researcher who resigned from Anthropic on September 10, publicly claimed that people inside AI development believe there is a greater-than-10% chance that advanced AI leads to human extinction within the decade.

Evan Hubinger, an alignment lead at Anthropic, echoed that figure, putting the risk of AI-caused human extinction in the next ten years above 10% as well.

The most striking claim came from Marcus Williams, a member of OpenAI’s safety team, who estimated a 70% chance of human extinction within three years if governments fail to implement meaningful regulatory measures.

Paul Christiano, appointed to OpenAI’s board in September 2026, added a more measured but still serious voice to the conversation, warning of a meaningful risk of catastrophic loss of control in the near term.

Why so many colleagues aren’t buying it

The skeptics pushing back are not fringe voices. Current and former employees across multiple top AI labs told the BBC they find these predictions exaggerated and unsupported by concrete evidence. Some described their reaction to the extinction-probability claims as amusement.

What makes the current moment different is that researchers are resigning publicly, posting probability estimates on social media, and giving interviews to national broadcasters. The argument has gone external, and that changes the stakes for everyone involved.

This internal conflict also sits within a broader industry context. In July 2026, more than 1,000 AI industry workers signed a petition calling for a slowdown in AI development, citing safety concerns.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.