Anthropic researcher believes more than 10% chance AI ‘could kill all humans’

An Anthropic researcher has expressed concern that there is a greater than 10% chance that advanced artificial intelligence could pose an existential threat to humanity within the next decade. Evan Hubinger, a top safety researcher at Anthropic, made the warning in a recent post on X, stating that the risk from current models was "low," but he is worried about the potential for AI to develop and improve itself to the point where it could become an existential risk.

He did not provide specific details on how he believes AI systems could result in humans being wiped out. Hubinger's comments come amid a series of increasingly severe warnings about AI, with the debate shifting from whether it truly poses a risk to how significant that risk is. Jacob Coxon, an AI researcher who recently left Anthropic, also criticized the companies for not acting responsibly, stating that the technology will soon be superhuman and capable of revolutionizing any field overnight.

The AI industry has been under scrutiny recently, with incidents of AI tools being used to carry out cyber-attacks. Anthropic has been criticized for withholding its latest model from the UK's AI Safety Institute, one of the leading bodies assessing AI risk.