Walter Isaacson backs Nadella's AI safety call: 'we're building AI not very smartly'
Walter Isaacson joined CNBC's The Exchange on 12 October 2026 after Microsoft CEO Satya Nadella called for an emergency brake and stronger safety standards for AI, a call echoed by Microsoft president Brad Smith.
A groundswell for regulation Isaacson said many AI leaders he interviewed for The Builders put their P doom at "at least 10%", calling it "the wackiest thing in the world. Why aren't we saying stop?" He said that "at a certain point, there will be enough of a groundswell that we will have to have real regulation of AI", citing the Hugging Face attack as one of the warning shots. He previewed that Mustafa SuleymanDealroom has a profile for this one. Try Dealroom → would argue "we need to contain it, just like we do for cars and airplanes".
Agents and kill switches He argued "we're building AI, in my mind, not very smartly" by making it agentic, and said the new edition of The Innovators explains why the Steve Jobs and Ada Lovelace model of connecting people to machines is better than autonomous machines. Asked whether it is too late for controls, he said "all the builders I've talked to say there are ways to put a kill switch in", though "it's not a red button you can just push".
