1,189
total controls
47
apply to AI agents
14
control families
4
trust boundaries
The Four Trust Boundaries
Zone 1 — Human Zone
Governance, authorisation, user intent. Controls: AC-3, AC-6, AU-2, AU-3, PM-9
Zone 2 — Agent Execution
Isolated environment. Controls: SC-39, CM-7, SC-4, SC-28, SI-3
Zone 3 — Enterprise Systems
IAM for non-human identities. Controls: IA-8, IA-2, AC-2, SC-7, SI-10
Zone 4 — AI Model Provider
Supply chain security. Controls: SA-12, SA-9, SR-3, SC-7, SC-8
Control Family Mapping
| Family | Controls | Zone |
|---|---|---|
| AC (Access Control) | AC-2, AC-3, AC-4, AC-5, AC-6, AC-17, AC-20 | 1, 3 |
| AU (Audit) | AU-2, AU-3, AU-6, AU-12 | All |
| CM (Configuration) | CM-2, CM-3, CM-7, CM-8 | 2 |
| IA (Identification) | IA-2, IA-4, IA-5, IA-8 | 2, 3 |
| SC (System & Comms) | SC-4, SC-7, SC-8, SC-12, SC-13, SC-28, SC-39 | 2, 3, 4 |
| SI (Integrity) | SI-3, SI-4, SI-5, SI-10 | 2, 3 |
| SR (Supply Chain) | SR-2, SR-3, SR-5 | 4 |
| SA (Acquisition) | SA-4, SA-9, SA-12 | 4 |
Key insight: AI agents aren’t a new category of risk. They’re a new surface for existing categories of risk. Leverage established frameworks rather than adopting new ones.
Implementation
- Inventory all AI agents, capabilities, data access, and trust boundary crossings
- Gap Analysis — Map existing controls against AI agent requirements
- Extend Controls — Update policies, configure IAM for non-human identities
- Monitor & Iterate — Establish continuous monitoring for behavioural anomalies
Need help implementing this?
First strategy session is complimentary. We typically respond within 4 hours.