Surveys of AI researchers who publish at top conferences show that, in a December 2024 poll of 1,580 respondents, the median estimate for the chance that 'future AI advances causing human extinction or similarly permanent and severe disempowerment of the human species' is 10%, while the median estimate for the chance that human-level AI turns out 'extremely bad (e.g. human extinction)' remains 5%, unchanged since 2016.
Forecasters with strong track records give lower numbers: the median superforecaster in mid‑2026 judged the chance of an AI‑caused catastrophe that kills more than 10% of people by 2100 at 2.4%, compared with a median of 5% among domain experts, and public concern is high, with a September 2026 YouGov poll finding 50% of US adults saying they are worried that AI will cause the end of the human race.
The image illustrates the progression of AI milestones and risk perception discussed in the article, matching the timeline of survey results from 2016 to 2026.
Individual experts quoted in the source offer a wide range of personal probabilities—from Geoffrey Hinton’s 0.1 estimate (which he calls a gut feeling) to Yann LeCun’s statement that he is 'not worried at all' about AI wiping out humanity, while Dario Amodei has suggested a 25% chance of things going 'really, really badly'—but the source emphasizes that these figures are subjective judgments, not measurements, and many researchers worry more about harms like deepfakes than about existential risk.