AIAGENT-004: Copilot agent access is scoped to security groups
- Platform
- Entra ID / M365
- Category
- AI Agent Governance
- Severity
- Medium
- Zero Trust pillar
- Identity (weight 2)
- Golden fixtures
- 3
- Branch coverage
- Observed: fixtures prove the verdicts they exercise
- Provenance
- baseline
What it checks
An authenticated Copilot Studio agent with no authorized security groups is reachable by any authenticated user in the tenant. Scoping an agent to specific security groups limits its audience to those with a business need, applying least privilege to who can invoke an agent and the data it fronts. This check flags authenticated agents that have no authorized security groups configured.
Recommended value
Authenticated agents that front sensitive data are scoped to specific authorized security groups rather than all authenticated users
Remediation
For agents that front sensitive data, configure authorized security groups in the agent's access settings so only members of those groups can interact with the agent. Leave tenant-wide access only for agents deliberately intended for all staff.
Fixture-proven verdicts
Every verdict below is proven by a golden fixture in the module's gating test suite. This table derives from the last green run; it cannot be edited by hand.
| Scenario | Expected verdict |
|---|---|
| clean | PASS |
| not-assessed | Not Assessed |
| unscoped | WARN |
Framework mappings
- NIST SP 800-53
- AC-2, AC-3, AC-6
Zero analytics. Zero tracking. Zero cookies. This site makes no external requests, so there is nothing to consent to.
Guerrilla is free and open source. · Validation
Guerrilla is written and maintained by Jim Tyler. This copy lives on jimtyler.org; the project also publishes at guerrilla.army.