Prove control coverage from how agents actually behave.

Your seat answers: what do agent behaviors mean for the frameworks we answer to, and what evidence do we have? Findings auto-map to controls — this is the evidence path, not a green checkmark. A framework with no findings is not 'passing'; it is unmeasured.

Bottom line

4 of 4 tracked frameworks have evidence; marketing-gpt drives most high findings (2 of 2). Every tracked framework has at least one finding behind it. This is contributing evidence — not a conformance claim.

Do nextReview framework coverage →Check sensitive-data exposure →

4/4

conformance coverage

frameworks with ≥1 finding (evidence)

high-severity findings

mapped to controls

marketing-gpt

drives most high findings

2 high findings

sensitive paths touched

across the fleet

Tracked frameworks — evidence vs. measurement gap

4 of 4 have evidence

Framework familyFindingsHighState

OWASP LLM52evidence on file MITRE ATLAS21evidence on file NIST AI RMF30evidence on file OWASP Agentic11evidence on file

A zero-finding framework is not a pass — it means we have no behavioral evidence either way. Treat it as a measurement gap to close, not a clean bill of health.

High findings by agent

accountability — who to chase

marketing-gpt

Your surfaces

Compliance & Frameworks4/4 covered

Findings mapped to the controls you answer for.

Data Protection6 paths

Sensitive files agents reached into.

Findings2 high · 3 med

Behavioral findings behind the controls.

Showback by User5 users

Attribute exposure to a person.

Conformance coverage = frameworks-with-evidence / frameworks-tracked. High findings grouped by f.agent. Derived from app/data/c16-fleet.ts.