These visualizations are experimental and still being refined. Designs and data presentation may change.
Incidents Over Time
How are incidents distributed over time? Color by risk subdomain or causal taxonomy dimension (entity, intent, timing), with a domain filter for cross-cutting drill-down.
Source: MIT AI Risk Repository
Key Takeaways
1.1,437 incidents classified across 24 risk subdomains from 2012 to present.
2.4.3 Fraud, scams, and targeted manipulation has the most incidents at 393 (27%).
3.Use the domain filter to drill down into subdomain distributions within each of the 7 risk domains.
4.Most domains show increasing concentration in 1-2 dominant subdomains over time.
Incident data from the AI Incident Database, classified using the MIT AI Risk Repository taxonomy. Incidents prior to 2012 not shown.
