Experts outline five most likely scenarios in which AI could actually wipe out all of humanity

From the Newshound archive · first published 23 September 2026.

An Anthropic staff member, Evan Hubinger, has publicly backed a former colleague's warning that AI could wipe out all humans by 2030, putting the odds at greater than one in ten. Jacob Coxon quit Anthropic after just four months this month, declaring the firm's own builders believe their technology could kill everyone within the decade. Finnish President Alexander Stubb secured signatures from 22 countries at the UN General Assembly demanding AI remain under human control — but the US, China and Britain all declined to sign.

AI researcher Toby Walsh ranks bioweapons and societal collapse among the most probable of five scenarios by which superintelligent AI could end humanity.

Trump now wants the technology rebranded 'Super Intelligence', arguing 'artificial' makes it sound fake — a move critics say only makes it more alarming.