The AI Risk Repository: A comprehensive database of risks from AI systems
As reported in numerous articles in the long-running RealKM Magazine artificial intelligence (AI) series, while there are clear benefits that can flow from the use of AI in knowledge management (KM), there are associated risks that must be considered and addressed. However, the lack of a comprehensive assessment of AI risks has hampered the shared understandings needed to plan and implement appropriate responses.
The new MIT AI Risk Repository addresses this gap. It should become an essential tool and reference for the KM community as it moves forward with AI. An overview of the repository is presented in the video below.
As documented in a preprint1, an evidence-based approach has taken to developing the AI Risk Repository involving a systematic review, forwards and backwards searching, and expert consultation. This identified 43 AI risk classifications, frameworks, and taxonomies, from which 700+ risks were extracted into a living AI risk database that can be easily accessed, modified, and updated.
The AI Risk Repository database can be filtered based on two overarching taxonomies, the Causal Taxonomy of AI Risks and the Domain Taxonomy of AI Risks. A best fit framework synthesis2 approach was used to create these taxonomies.
The Causal Taxonomy of AI Risks classifies each risk by its causal factors:
- Entity: Human, AI.
- Intentionality: Intentional, Unintentional.
- Timing: Pre-deployment; Post-deployment.
The mid-level Domain Taxonomy of AI Risks classifies risks into seven AI risk domains:
- Discrimination & toxicity.
- Privacy & security.
- Misinformation.
- Malicious actors & misuse.
- Human-computer interaction.
- Socioeconomic & environmental.
- AI system safety, failures, & limitations.
These are further divided into 23 subdomains.
Acknowledgements: With thanks to Elizabeth (Liz) McLean, MSLS CKS for alerting me to the release of the AI Risk Repository.
Header image source: AI Risk Repository.
Reference:
- Slattery, P., Saeri, A. K., Grundy, E. A., Graham, J., Noetel, M., Uuk, R., … & Thompson, N. (2024). The AI Risk Repository: A Comprehensive Meta-Review, Database, and Taxonomy of Risks From Artificial Intelligence. arXiv preprint arXiv:2408.12622. ↩
- Carroll, C., Booth, A., Leaviss, J., & Rick, J. (2013). “Best fit” framework synthesis: refining the method. BMC medical research methodology, 13, 1-16. ↩
