Researchers at the Massachusetts Institute of Technology (MIT) have published the world’s first comprehensive database, designed to document risks associated with AI. The AI Risk Repository was compiled and published online by the FutureTech group at MIT’s Computer Science & Artificial Intelligence Laboratory. The database documents over 700 potential risks that advanced AI systems could pose. It is understood to be the most comprehensive source of information yet about previously identified issues that could arise from the creation and deployment of AI models. Just 10% of AI risks are detected before deployment, the AI Risk Repository said, underscoring the importance of post-launch monitoring. To develop the AI Risk Repository, the FutureTech team used peer-reviewed journal articles and preprint databases detailing AI risks. The findings showed that the most common risks were related to AI system safety and robustness (76%), unfair bias and discrimination (63%), and compromised privacy (61%).
The #AI Risk Repository reveals that only 10% of AI risks are detected before deployment, underlining the importance of post-launch monitoring.@MIT https://t.co/qZcyrcSUTl
— Regulation Asia (@RegulationAsia) August 20, 2024
