Mission

AI agents are being deployed at unprecedented scale. Organizations have governance frameworks for AI strategy, risk, and compliance. What they do not have is a framework for governing what those agents actually do in production — their behavior, their failure modes, the patterns that recur across deployments.

AI behavioral governance is that discipline. This initiative exists to build it: an open incident database, a behavioral pattern taxonomy, and a governance methodology derived from real production operations. Independent. Evidence-based. Practitioner-facing.

Mission statement

To define and advance the discipline of AI agent behavioral governance — the systematic study of how agents fail under delegated authority, and the methodology for preventing those failures at scale.

AI behavioral governance is not AI governance

AI behavioral governance is distinct from AI governance. AI governance addresses board oversight, risk committees, ethics policies, and regulatory compliance — the organizational structures responsible for AI. AI behavioral governance addresses what AI agents do once operating in production. Both are necessary. Neither substitutes for the other. This initiative focuses exclusively on the behavioral governance layer.

About the researcher

This site is maintained by John McCormick, an AI systems architect and independent researcher. The behavioral pattern taxonomy was developed through direct observation of AI agent failures in production — not through theoretical threat modeling, but from operating and governing multi-agent systems and finding that standard incident frameworks could not classify what was failing or why. The full governance framework is published at aiagentgovernance.org. The taxonomy was formally submitted March 2026 under NIST docket NIST-2025-0035.

McCormick can be reached through the contact page or on LinkedIn.

Case studies

The governance framework is documented in principle at aiagentgovernance.org. The following case studies show it operating under real conditions — actual audit trails, actual defects, actual remediation cycles.

• Routing Plan Enforcement — Full Governance Lifecycle — A complete governance lifecycle trace: specification through final approval, including two remediation loops and a behavioral incident detected mid-execution.