Skip to main content
This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
MIT AI Risk Initiative
These visualizations are experimental and still being refined. Designs and data presentation may change.

Cross-Dataset Network

A force-directed map of every risk, incident, governance document, and mitigation in the repository, connected to the 46 risk subdomains they touch. Clusters reveal dense coverage; isolated hubs reveal gaps.

Multi-connection only5,300 nodes · 10,158 connections
Discrimination & ToxicityPrivacy & SecurityMisinformationMalicious Actors & MisuseHuman-Computer InteractionSocioeconomic & Environmental HarmAI System Safety, Failures, & Limitations|RisksIncidentsGovernanceMitigations V2

Click a record to open its details. Click a subdomain hub for its summary. Drag to explore, scroll to zoom, click the background to re-center.

Key Takeaways

  • 1.Capability & robustness is the densest subdomain, with 1,043 connections across risks (126), incidents (306), governance (486), and mitigations (125).
  • 2.AI welfare is the least represented, attaching to only 9 records across all four datasets — a likely coverage gap.
  • 3.Governance account for the largest share of connections at 55% (5,561 of 10,158 edges).
  • 4.Environmental harm shows a notable imbalance: 57 risks but only 6 mitigations — risks outpace documented responses.