AI could kill us all within a decade, top researcher warns

A senior safety researcher at Anthropic has issued a stark new warning, stating there is now a greater than 10 percent chance that advanced AI systems could wipe out humanity within the next ten years.

Evan Hubinger acknowledged that current models pose only a "low" risk, but expressed deep concern that the technology may soon reach a point of self-improvement that transforms it into an existential threat.

While he did not specify exactly how such a catastrophe might unfold, his comments add to a growing chorus of experts shifting the debate from whether AI poses a danger to just how severe that danger truly is.

Hubinger's remarks came in response to a social media post from Jacob Coxon, a former Anthropic and OpenAI researcher who recently left the company.

Coxon accused both firms of acting irresponsibly, warning that "superhuman systems" are imminent.