These visualizations are experimental and still being refined. Designs and data presentation may change.

High Severity Incidents

Filter incidents by harm severity threshold to identify which risk domains concentrate the most serious real-world harms. Bars are stacked by domain, with controls to adjust the severity cutoff, focus on a specific harm category, or switch between count and proportion views.

Discrimination & Toxicity

Privacy & Security

Misinformation

Malicious Actors & Misuse

Human-Computer Interaction

Socioeconomic & Environmental

AI System Safety, Failures & Limitations

Key Takeaways

1.460 incidents have at least one harm category scoring ≥3 (medium or higher).
2.Malicious Actors & Misuse is the most represented domain among high-severity incidents at 37%.
3.Toxic or Malicious Content is the most common high-severity category with 187 incidents.
4.High-severity incidents are concentrated in recent years, reflecting both increased AI deployment and improved incident tracking.