This directory contains the full ACR (Agentic Control at Runtime) Standard documentation—a reference architecture for governing autonomous AI systems at runtime.
Choose a path based on your role. Each path is 3–5 steps to the most relevant docs.
| Role | Goal | Path |
|---|---|---|
| Architect | Design runtime governance into your AI systems | Control Plane Architecture → Runtime Architecture → Implementation Guide → Telemetry Schema |
| Security | Understand threats and controls | STRIKE Threat Model → Behavioral Policy Enforcement → Self-Healing & Containment → Identity & Purpose Binding |
| Compliance | Map ACR to frameworks and audits | NIST AI RMF Mapping → Execution Observability → Human Authority → SOC 2 / ISO placeholders |
| Implementer | Build or integrate ACR-aligned controls | Implementation Guide → Use Cases → Telemetry Schema → Pillars |
| Evaluator | Assess adoption or vendor alignment | Adoption & Maturity → Pillars Overview → STRIKE Summary |
| Section | Description |
|---|---|
| Pillars | Overview of the six control layers |
| Architecture | Control plane, runtime, and production lifecycle |
| Specifications | Telemetry schema and future specs |
| Compliance | NIST AI RMF and other mappings |
| Security | STRIKE threat model and glossary |
| Guides | Implementation guide, use cases, FAQ |
- Identity & Purpose Binding
- Behavioral Policy Enforcement
- Autonomy Drift Detection
- Execution Observability
- Self-Healing & Containment
- Human Authority
- Telemetry Schema — JSON schema for observability events
- Implementation Guide — Deployment patterns and reference architectures
- STRIKE Threat Model — AI-specific runtime threats
- NIST AI RMF Mapping — Compliance alignment
- FAQ — Frequently asked questions
- Adoption & maturity levels — How to adopt and claim alignment