Anthropic researcher warns more than 10% chance AI ‘could kill us all’

A top safety researcher at Anthropic has resigned, warning artificial intelligence is advancing so quickly he believes there is a greater than 10% chance it “could kill all humans” by the end of the decade. Posting on X, Evan Hubinger said the risk from the models which currently exist was “low”, but that the technology might develop into “superhuman systems that can hack anything” and “acquire real power and resources”. We hear from the computer scientist often described as ‘Godfather of AI’.

Also in the programme: Israel announces further measures against the UK in response to sanctions imposed on goods from settlements in the occupied West Bank; new research suggests the consumption of very hot drinks might increase the risk of a type of oesophageal cancer; and does a hot, dry summer mean more alcoholic champagne?

(Photo: illustration show Anthropic logo, a keyboard and a robotic hand, June 5, 2026. Credit: Reuters/ Dado Ruvić)