What changes
Before and after
The classic way
- Policy is a PDF; evidence is gathered once a year for the audit.
- Security reviews the system after it is built and files findings the roadmap never reaches.
- Nobody has a list of which agents exist, what they can reach and who approved them.
The AI way
- The standards now name AI software. NIST published SP 800-218A in July 2024, a community profile of the Secure Software Development Framework for generative AI and dual-use foundation models, and released the initial public draft of SSDF 1.2 (SP 800-218r1) for comment in December 2025 [4].
- The agent-specific risks have a list: the OWASP Top 10 for Agentic Applications, covered test by test in the test center and incident by incident in the Learning Center [11].
- Evidence is produced by the pipeline as it runs: provenance from the build [5], the hook log from the agent's session [9], the approval from the environment gate [10], and an audit trail of every agent action, which the Learning Center lists as a weakness when it is missing.
Step by step
The three steps of Govern
- 1
Policy becomes code Person
Security writes the rules as policy code and hooks: what an agent may never run, which files it may never touch, which egress is allowed. The rules are reviewed like any code change.
- 2
The pipeline produces the evidence AI
Each run leaves attestations, logs and approvals behind it. An agent assembles them against the controls of a framework such as the SSDF, so the audit reads the pipeline rather than interviewing the team [4].
- 3
People own the exceptions Gate
An exception to a policy is a pull request with a named approver and an expiry. The loop closes: what governance learns here becomes the next stance in Plan, and the lifecycle starts again.
The gate
An inventory of agents and their permissions, an exception process with names and expiry dates, and an audit trail that can replay any agent's actions.
What the AI does here
Maps evidence to controls, drafts the exception request, watches for drift between policy and the pipeline, and flags an agent whose permissions grew.
Tools people use at this stage
- NIST SSDF and SP 800-218A
- OWASP Top 10 for Agentic Applications
- OPA and Conftest
- GitHub rulesets and audit log
Named as examples, not endorsements; the center is not affiliated with any of them.
What this stage adds
The risks, and the practices that answer them
Each risk is a weakness in the Learning Center, with the incidents where it happened; each practice has its sources there.
Risks
- Missing audit trail of agent actionsAgent actions leave no audit trail.
- Identity and privilege abuse by agentsAn agent's identity and privileges grow unnoticed.
- Overly broad OAuth scopes for AI connectorsA connector holds OAuth scopes nobody reviewed.
Read for this stage
Sources
- [4] NIST: Secure Software Development Framework, news (SP 800-218A, draft SSDF 1.2)read 2026-10-27
- [11] AI Test Automation: every incident earns a test (this site)
- [5] SLSA v1.1: security levelsread 2026-10-27
- [9] Claude Code docs: automate actions with hooksread 2026-10-27
- [10] AI Test Automation: the AI SDLC on GitHub (this site)