These 5 AI risks have the highest potential for catastrophe
TL;DR
Researchers from MIT and the University of Queensland surveyed 272 international experts on 24 AI risks for the period from 2025 to 2030. They put the odds at one in five that AI gains dangerous weapons capabilities or causes mass harm killing millions within the next five years. Experts rated 18 of the 24 risks as carrying at least a 10% probability of catastrophic outcomes.
Nauti's Take
Numbers like these are less a reason to panic than an opportunity: they make risks comparable and give policymakers and companies a basis for setting priorities instead of gut feeling. The catch is the method — expert estimates across a five-year horizon are opinion, not measurement, and the spread stays wide.
Anyone running AI in production should follow the debate but anchor decisions in concrete security practices rather than percentages.
Briefingshow
Until now the AI safety debate has run largely on anecdotes and isolated incidents. A ranked list with probabilities moves the discussion into a format regulators, insurers and compliance teams can actually process, and one that gets cited when rules for frontier models are negotiated. For companies deploying AI, the practical consequence is that they will increasingly be asked which of these risk categories they have reviewed.
Sources