AI Boss Tools | Diagnostic 04 • Back to the Hub
MIT's AI Risk Repository catalogues more than 1,700 documented AI risks and sorts them into seven domains. This tool walks you through all seven for your own organisation: you rate the likelihood and the impact of each, and it plots your risk matrix with a rating and a next step for every domain.
A risk matrix is the oldest tool in risk management for a reason: it forces you to separate how likely something is from how badly it would hurt. Most organisations have never applied one to AI, partly because nobody agreed on what the risks were.
MIT solved the naming problem. The AI Risk Repository, first published by Slattery and colleagues in 2024 and updated since, reviewed 74 risk frameworks and produced a single taxonomy: seven domains and 24 subdomains covering everything from discrimination and privacy to misinformation and system failure.
You rate each domain twice: likelihood (1 to 5) and impact (1 to 5). Likelihood times impact gives a score out of 25, rated Low, Moderate, High, or Critical. The one-pager gives you a board-ready summary of all seven.
A living database of more than 1,700 AI risks drawn from 74 frameworks, classified into seven domains and 24 subdomains. The domain structure of this tool follows it directly.
The peer-reviewed meta-review behind the repository, from MIT FutureTech and collaborators.