These three attacks show how an adversary can subvert an autonomous agent at different points in its pipeline: by planting a goal in the agent's context (agenda seeding), by tricking the agent into binding to a malicious tool provider, or by exploiting the way separate components compose together. Read together they reveal that the trust gaps in agentic systems are layered — the input, the tools, and the composed chain can each be abused.
Glossary topic
Agentic AI attack chain
Terms in this topic
Track this in the live feed
See how this plays out in real AI security and governance developments.
Open the feed →