Standards and guidance
OWASP Agentic Top 10: Risk & Control Reference
A risk taxonomy and mitigation resource for agentic applications, rather than a certification scheme.
How to use this reference
Use the relevant risk categories to identify test scenarios and control gaps. Applying a taxonomy does not establish that every risk has been tested.
Evidence to discuss
Identify the system scope, control owner, implementation evidence, test results, exceptions and the reviewer responsible for acceptance. Requirements are mapped to the agreed source version; missing and untested controls remain visible.
Source and status
OWASP official agentic applications release
Reference checked on 7 October 2026. This page is an engineering overview, not legal advice or a claim of accreditation, partnership or endorsement.
Using a risk taxonomy in a review
These are engineering discussion prompts, not a substitute for the source requirements or the appointed reviewer's criteria.
- Which risk categories relate to the agent's actual tools and actions?
- What scenario would exercise each relevant control boundary?
- Which categories were excluded, and why is that exclusion reasonable for this deployment?
Use the security review checklist to organise owners, test records and gaps. The assessment method explains how coverage limits are recorded.
Start with a clear scope
Tell us which systems, actions and review requirements are in scope. We will discuss the work, responsibilities and deliverables before you commit.