Microsoft Moves AI Governance From Policy to Runtime Enforcement
Microsoft shifts AI governance from the policy layer to runtime: nine domains, four functions, continuous enforcement as agents operate in production.
Why it matters
As agents move into production workflows, governance can't be a pre-deployment checklist — it has to happen live. Microsoft's architecture (policy + runtime enforcement + continuous eval + audit) is the infrastructure that makes agentic deployments auditable and compliant at scale.
The key facts
12 to knowNine governance domains outlined
Four governance functions: policy, control, visibility, proof
Runtime enforcement (not just pre-flight policy)
Continuous evaluation and observability during operation
Identity, security, and audit evidence as first-class components
Designed for production agents and AI applications
Published August 2026 — current architecture guidance
Four core functions: policy, control, visibility, proof
Runtime enforcement tied to policy
Continuous evaluation and observability as core components
Identity, security, and audit evidence integrated
Focus on production agent operation and verification
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Microsoft has outlined an AI governance architecture spanning nine governance domains and four functions: policy, control, visibility, and proof. The approach connects policies with runtime enforcement, continuous evaluation, observability, identity, security, and audit evidence to help…