AI-related types of incidents reported by organizations in the past two years

According to the Artificial Intelligence Index Report 2025, organisations have reported a rise in AI-related incidents over the past two years.

Adversarial attacks (56%) and privacy violations (55%) are the most prevalent, highlighting vulnerabilities in AI systems. Unintended decision-making (51%) and model bias (47%) underscore ethical concerns, while performance failures (46%) reveal technical shortcomings.

The AI Incident Database (AIID) tracks these issues, emphasising the need for robust safeguards, transparency, and accountability in AI deployment.

As AI integration expands, addressing these risks is critical to ensuring responsible use and maintaining public trust in emerging technologies.