The AI Model Risk Catalog: What Developers and Researchers Miss About Real-World AI Harms
Not on the map. Why?
Abstract (arXiv)
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 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.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 26 of its 72 references that are in the map. First posted on arXiv 21 Aug 2025.
Builds on
26 of 72 references.
References · 26
- 2019Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AIcited by 1,552
- 2018Model Cards for Model Reportingcited by 1,080
- 2021Ethical and social risks of harm from Language Modelscited by 966
- 2023Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignmentcited by 401
- 2023Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Biascited by 232
- 2023Stable Bias: Analyzing Societal Representations in Diffusion Modelscited by 226
- 2021A survey on datasets for fairness-aware machine learningcited by 157
- 2023Sociotechnical Safety Evaluation of Generative AI Systemscited by 157
- 2022Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reductioncited by 138
- 2024Bias and Unfairness in Information Retrieval Systems: New Challenges in the LLM Eracited by 98
- 2020Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Databasecited by 96
- 2022Interactive Model Cards: A Human-Centered Approach to Model Documentationcited by 61
14 more
- 2024Farsight: Fostering Responsible AI Awareness During AI Application Prototypingcited by 28
- 2023Digital Deception: Generative Artificial Intelligence in Social Engineering and Phishingcited by 27
- 2023AHA!: Facilitating AI Impact Assessment by Generating Examples of Harmscited by 26
- 2025Lessons From Red Teaming 100 Generative AI Productscited by 23
- 2022Aspirations and Practice of Model Documentation: Moving the Needle with Nudging and Traceabilitycited by 14
- 2024ExploreGen: Large Language Models for Envisioning the Uses and Risks of AI Technologiescited by 12
- 2024AI Cards: Towards an Applied Framework for Machine-Readable AI and Risk Documentation Inspired by the EU AI Actcited by 10
- 2024Co-designing an AI Impact Assessment Report Template with AI Practitioners and AI Compliance Expertscited by 10
- 2025RiskRAG: A Data-Driven Solution for Improved AI Model Risk Reportingcited by 9
- 2025Atlas of AI Risks: Enhancing Public Understanding of AI Riskscited by 7
- 2021Social Engineering in Cybersecurity: A Domain Ontology and Knowledge Graph Application Examplescited by 5
- 2021"This Whole Thing Smacks of Gender": Algorithmic Exclusion in Bioimpedance-based Body Composition Analysiscited by 3
- 2024The Atlas of AI Incidents in Mobile Computing: Visualizing the Risks and Benefits of AI Gone Mobilecited by 2
- 2024Effective Mitigations for Systemic Risks from General-Purpose AIcited by 2
Built on by
Citing papers per year: 2025: 1; 2026: 2.
3 papers.