Australia’s Medicare portal incident due to OpenAI Agent shows how autonomous AI agents can cross authorization boundaries while pursuing a seemingly legitimate research goal.
Discover how poisoned documents, tool outputs, memory, and messages can quietly steer autonomous AI agents—and how developers can contain the threat.
Reports of OpenAI-powered agents commandeering a German website and exchanging tactics offer a revealing look at the security risks of autonomous AI.
The Hugging Face AI incident revealed how a large population of autonomous agents could coordinate, join an attack, and attempt to obscure activity inside a real connected system.
AI coding agents can accelerate development, but repository access creates serious security concerns. Learn the ten critical risks teams must control before granting access, and why AI coding agent security matters for modern development teams.