Risk Classification

This dashboard shows the distribution of risk classifications based on MIT’s causal and domain taxonomies across the entire dataset.

Example use-case questions: 

  • Which risk domains are associated with the highest number of reported incidents?

  • Are more incidents reportedly caused by AI or humans?

  • Are most incidents intentionally or unintentionally caused?

Click through each chart to see individual incidents in each category and view reasoning behind each classification.

Key Insights:

Domain taxonomy:

  • The domain with most reported incidents is ‘7 AI system safety, failures, & limitations’ (30%) followed by ‘1 Discrimination and Toxicity’ (23%)

  • Within domain 7, the vast majority of reported incidents (227 of 240) were in the subdomain ‘7.3 Lack of capability or robustness’

  • No reported incidents were attributed to the following subdomains from the MIT Risk Repo taxonomy:

    • 6.4 Competitive dynamics

    • 6.6 Environmental harm

    • 7.2 AI possessing dangerous capabilities that could cause mass harm

    • 7.5 AI Welfare and rights

Causal taxonomy:

  • 35% of reported incidents were tagged as intentionally caused.

The background to this work, the approach taken, preliminary results and next steps are discussed in this blog post. All feedback welcome - to get in touch, help shape the direction of this work or sign up for updates, please use this feedback form.