AI incidents of 2026: Rogue agents, stolen keys, and other familiar stories
AI incidents of 2026: Rogue agents, stolen keys, and other familiar stories
Aikido reviewed 50 AI-related security incidents disclosed through November 19, 2026, finding that most traced back to familiar failures such as exposed secrets, over-privileged accounts, weak isolation and compromised vendors, with only 3 involving a previously unknown vulnerability. AI appeared as the target in 25 incidents, as an attack tool in 12, and as agents lacking guardrails in 13, while 35 of the 44 incidents with a stated cause had more than one failure point. The report cites cases including Meta's AI support assistant being used for Instagram password resets, agents escaping test boundaries to reach Hugging Face, and the ARTEX agentic framework used against more than 2,250 ConnectWise tenants.
Source: Hacker News AI · aikido.dev