Published in December 2025, this list (ASI01 to ASI10) covers risks that appear once a model can plan and act. Examples are agent goal hijack, tool misuse, identity and privilege abuse, memory poisoning, and rogue agents. It builds on the earlier OWASP agentic threat taxonomy, so read the two together.
Read in this order
- OfficialOWASP Top 10 for Agentic Applications for 2026Start here: the ten agent risks, each with a plain definition and examples.
- OfficialAgentic AI: Threats and MitigationsThe earlier threat taxonomy the list builds on, with mitigations per threat.
- OfficialSecuring Agentic Applications Guide 1.0Turns the threats into build and deploy controls you can apply to a real agent.