Categories
Incidents in the Index are classified into five primary categories that reflect the nature of the harm or failure documented.
Data exposure
Incidents where confidential, personal, or proprietary information was unintentionally disclosed or made accessible through AI systems.
/categories/data-exposure
Hallucination
Cases where AI systems produced factually incorrect, fabricated, or misleading outputs that caused measurable harm or material misrepresentation.
/categories/hallucination
Bias and discrimination
Cases where AI systems produced unfair or discriminatory outcomes toward individuals or groups based on protected characteristics.
/categories/bias-discrimination
Deepfakes
Incidents involving AI-generated synthetic media used to deceive, defame, harass, or defraud with documented harm.
/categories/deepfakes
Safety failure
Cases where an AI system caused, contributed to, or failed to prevent physical harm or serious safety risk.
/categories/safety-failure