MIT and Queensland University Study Shows 12% Risk of AI-Enabled Bioweapons Causing Catastrophic Harm by 2030

According to Axios, a joint study by MIT FutureTech and the University of Queensland released today shows that 272 researchers assessing 24 AI risks identified a 12% probability that AI capabilities could lead to catastrophic outcomes—defined as over 1 million deaths or more than $100 billion in losses—before 2030 with current mitigation measures in place. The same probability applies to AI-enabled weapons and mass harm. Without mitigation measures, this risk exceeds 20%. The research ranked AI-assisted chemical and biological weapons construction among the highest-risk categories.

Multiple leaders from major AI companies have also issued a joint public letter warning of AI-derived bioweapon risks, calling for stronger model capability audits and safety safeguards, particularly addressing the potential misuse of open-source models.