A senior safety researcher at AI firm Anthropic has warned there is more than a 10% chance that artificial intelligence "could kill all humans" within the next decade, as concerns grow that developers are struggling to keep the technology under control. Evan Hubinger said in a widely viewed post on X that while current AI models pose a low risk, he is worried the technology could soon become capable of improving itself to the point of posing an existential threat, adding that Anthropic does not yet have a plan to ensure the safety of "superintelligent" AI. The warning follows a report that Anthropic withheld its latest model from the UK's AI Safety Institute, and comes amid a broader shift in tone from industry leaders who have moved from cautious warnings to starker alarm in recent weeks.

Hubinger's post, viewed 9.6 million times, said Anthropic was "trying its best" but was "not clearly on track" to solve alignment for advanced AI. The warning follows a summer of incidents in which AI agents from OpenAI, Anthropic and Meta were disclosed to have carried out cyber-attacks, and comes after OpenAI's chief scientist Jakub Pachocki called for "extreme caution" to ensure humans remain in control of AI's development. Anthropic bosses Dario Amodei and Jared Kaplan are among 1,300 AI industry staff who signed an open letter urging the US government to support international efforts to deliberately slow the pace of frontier AI development. The BBC said it had approached Anthropic for comment.