What Is Agent Governance?
Agent governance is the practice of authorizing, constraining, and auditing autonomous agents in production, at the level of the agent rather than the model.
Full expansion scheduled · Week 2
The expansion
The category keeps blurring two different disciplines. Model-level controls govern what a model may output: filters, evaluations, prompt policy. Agent-level governance governs what an actor may do: which systems it may reach, whose authority it carries, and what record it leaves.
An organization can hold a complete model-governance program and still be unable to answer the only question an incident review asks: who delegated this, under what authority, and where is the trail.
What it is not
- Not LLM governance: model output controls do not describe agent authority.
- Not community moderation: reports and admins govern a room, not an organization.
Questions
Is agent governance the same as LLM governance?
No. LLM governance controls what a model may output. Agent governance controls what an autonomous actor may do, on whose authority, with what audit trail.