AI Extinction Risk Warning

On March 27th, Kokotajlo — who left OpenAI last year and has since become one of the more credible internal critics of the organisation's safety culture — appeared to tell The Daily Show's host that he estimates a 70% probability of human extinction from AI within approximately five years. The host's visible shock was, understandably, the part that got clipped and circulated. Kokotajlo is not a fringe voice: he was among the signatories of public letters warning about AI risk and left OpenAI citing safety concerns, which gives his estimates a weight that a random commentator's would not carry. That said, a single television appearance generating two near-identical signals on the same day is the signature pattern of a story that has travelled on heat rather than substance.

If confirmed, here is what this means. A credentialed AI insider putting a 70% extinction probability on a five-year timeline — publicly, on mainstream television — marks a genuine shift in how catastrophic risk is being communicated. Not in academic papers or niche podcasts, but in the living rooms of people who have never heard of alignment theory. The second-order effect is political pressure: legislators who can ignore an arXiv preprint cannot as easily ignore a former OpenAI researcher saying this to a late-night audience of millions. For the AI industry, it accelerates the uncomfortable question of whether voluntary safety commitments are any longer a credible substitute for regulation.

Watch for any written follow-up from Kokotajlo himself — a Substack post, an interview with a technical outlet — that either substantiates the 70% figure with reasoning or walks it back. That will tell you whether this was a considered public statement or a number that ran away from its context.

NewsHive monitors these sources continuously. All signal titles above link to the original reporting.

Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →