Agentic Trust Controls: 65 Open-Source Controls to Govern AI Agents

Agentic Trust Controls is an open-source framework that extends ISO 27001 and ISO 42001 to address the unique risks of AI agents. It offers 65 controls across 12 domains, with two usage baselines: a developer baseline (43 controls) for building agents, and a user baseline (22 controls) for deploying them. The controls cover identity, action guardrails, memory protection, adversarial testing, and more, providing a practical starting point for teams that lack guidance.