MIT warns: AI risks remain high, even after mitigation
The MIT AI Risk Initiative has published a comprehensive Delphi study in which 272 experts assessed 24 AI risk categories based on probability, impact, and societal consequences.
The main conclusion: prioritizing risks is becoming crucial, as not all AI risks can be effectively addressed simultaneously. Notably, all 24 risk categories maintain at least a 5% probability of a catastrophic outcome before 2030, even after mitigating measures.
For 18 of the 24 risks, that probability is even higher than 10%, including unemployment, inequality, loss of privacy, and the erosion of democratic values.
The study emphasizes that organizations should not only focus on compliance but especially on residual risks: even a small percentage chance of catastrophic damage can be socially unacceptable.