Autonomous Agent Audit
An Autonomous Agent Audit tests your AI agent for security vulnerabilities, safety risks, and quality issues using autonomous red teaming agents. It explores how your agent behaves, builds a risk model for its use case, and runs targeted attacks to find weaknesses.
Use the results to understand how failures could affect your application, prioritize fixes, and build regression tests.
What you get
Section titled “What you get”- AI Trust level: An overall grade and assessment of your agent’s behavior.
- Risk model: The risks identified for your agent and the scenarios tested.
- Findings and recommendations: Identified issues, supporting evidence, and proposed fixes.
- Reusable tests: Generated attack scenarios and results that you can review and use in evaluations.
Start with a Light audit for a quick first look at a prioritized subset of risk scenarios, and then run a Full audit to test all scenarios in the generated risk model across multiple attack waves.
How an audit works
Section titled “How an audit works”A Light audit runs one attack wave, while a Full audit runs multiple waves.
graph LR
I[Input evaluations] -.->|optional| A
A([Reconnaissance]) --> B([Risk modeling])
B --> C([Attack wave])
C -->|"Next wave (Full only)"| C
C --> D([Report generation])
Audit report
Section titled “Audit report”The PDF report brings together the priority findings, their severity, and the AI Trust assessment. Download the report to share the results with your team.
