Risk test
The Three-Circle Risk Test
Score any AI use case in seconds. Risk lives in the intersection, not in any single factor. Answer the three questions for the system you are building.
1Think about what you're building
The test applies to anything on the maturity model — pick the closest as your frame of reference, then answer for your specific implementation.
2Answer three questions about your system
3Your risk
Low risk
0 of 3 — Low risk
No high-risk intersection. Standard controls for the capability are enough — keep least privilege and logging in place.
Pick what you’re building above to see recommended controls.
Adapted from Meta AI, "Agents Rule of Two: A Practical Approach to AI Agent Security" (Oct 2025), as presented in the AWS AI Security Framework. When all three apply, use deterministic controls — not probabilistic ones.