The AI Model Risk Catalog: What Developers and Researchers Miss About Real-World AI Harms
This work addresses the problem of incomplete risk reporting in AI development for developers and researchers, highlighting gaps in understanding human-interaction risks, though it is incremental in building on existing risk repositories.
The authors analyzed nearly 460,000 AI model cards to identify around 3,000 unique risk mentions, creating the AI Model Risk Catalog, and found that developers focus on technical issues like bias while overlooking fraud and manipulation, which are common real-world harms.
We analyzed nearly 460,000 AI model cards from Hugging Face to examine how developers report risks. From these, we extracted around 3,000 unique risk mentions and built the \emph{AI Model Risk Catalog}. We compared these with risks identified by researchers in the MIT Risk Repository and with real-world incidents from the AI Incident Database. Developers focused on technical issues like bias and safety, while researchers emphasized broader social impacts. Both groups paid little attention to fraud and manipulation, which are common harms arising from how people interact with AI. Our findings show the need for clearer, structured risk reporting that helps developers think about human-interaction and systemic risks early in the design process. The catalog and paper appendix are available at: https://social-dynamics.net/ai-risks/catalog.