The category KYE Protocol™ defines.
When humans and AI agents share authority over an action, the question is no longer “is the model well-behaved?” — it is “who held authority, within what purpose and scope, and can that be proven after the fact?” Agentic Governance™ is the discipline of resolving that question to finality. It governs the act, not merely the model. KYE Protocol™ is its reference implementation.
What the category covers
Agentic Governance™ is governance FOR agentic systems — any system where authority to act is delegated across a chain of humans, AI agents, and the tools they invoke. Its primitives are an act (something done), an authority (the delegated permission that licensed it), a purpose and a scope (the bounds the authority was admitted within), and an evidence chain that resolves the whole to finality. The essence: every act has authority; every authority has finality.
You hold four primitives. Each maps to a signed schema and replays from public keys alone — so a third party validates the decision against NIST 800-207, ISO/IEC 42001, and SOC 2 control objectives without trusting the vendor:
- Authority Resolution™ — tracing who held authority for an act across the delegation chain.
- Authority Finality™ — a decision that is sealed, signed, and replay-derivable from public keys alone.
- Evidence Finality™ — the evidence pack that makes the resolution contestable and auditable.
- Hybrid authority — humans-with-AI-agency, not just autonomous agents in isolation.
How Agentic Governance™ differs from adjacent categories
Three adjacent categories solve real but narrower problems. Agentic Governance™ sits at a different altitude: it governs the authority behind an act, and the others feed into it rather than replace it. The boundary below is exact — each category is admitted as an input, never as a substitute:
- vs AI-governance — AI-governance asks whether a model is fair, safe, and compliant in its behaviour. Agentic Governance™ asks whether the act the model drove was authorised, and proves it. Model behaviour is an input to the authority decision, not the decision itself.
- vs model-risk — model-risk (SR 11-7 lineage) validates a model statistically — performance, drift, explainability. Agentic Governance™ governs the runtime exercise of authority, regardless of which model produced the action.
- vs prompt-firewalls — prompt firewalls screen the prompt as input for injection and jailbreaks. Agentic Governance™ treats the prompt as an authority act — a firewall verdict can feed the governance decision, it does not replace it.
Define the category with us.
Agentic Governance™ is the organising principle of KYE Protocol™, surfaced through KYE GovernedUI™ and proven by a signed Evidence Pack™. The Open Agent Governance Foundation™ is the open venue for standardising it.