When AI fAIls, it lands here.

A daily tally of documented AI incidents and hazards: biased models, deepfake scams, cars that drive themselves into trouble. The numbers come from the three most serious public databases and refresh every morning.

Loading the latest numbers…

Fresh off the wire

Full wire →

fAIlcharts

stats.json

Fig. 1Documented incidents per year

Source: OECD AI Incidents Monitor. Current year shown to date, in red.

Fig. 2Monthly count, trailing 24 months

Source: OECD AI Incidents Monitor. Most recent month is partial.

Fig. 3What gets harmed

Source: OECD AIM harm types, all time. One incident can carry several harm types.

Fig. 4Risk domains

Source: MIT AI Risk Repository classification of AI Incident Database cases.

Hall of fAIl

All entries →