Non-profit Database Tracks Over 600 Real-World AI Incidents to Inform Developers and Policymakers
-
Non-profit AI Incident Database tracks and catalogs harms and risks from AI systems, with over 600 incidents logged so far
-
Funded by Underwriters Laboratories, modeled after cybersecurity and transportation safety incident reporting
-
Focuses on real-world impacts of AI, not just bugs and vulnerabilities
-
Mainly collects incidents via media reports and Twitter, with a recent surge due to generative AI issues
-
Aims to inform developers and policymakers to improve models and regulate pressing risks revealed by incident patterns over time