Agentic AI Governance, defined

Working definition. Agentic AI governance is the discipline of specifying, imposing and verifying duties — not rights, not intentions — on autonomous software agents. Its distinctive feature is that the duties are machine-readable and enforceable at execution time, not written for humans to interpret afterwards.

Why the shift from rights to duties

Legal frameworks written for humans use rights as the primary unit because humans have moral standing. AI agents do not. Governance frameworks that borrow the "rights" vocabulary from human ethics fail at the runtime level: no agent can enforce a right it does not itself understand, and no operator can audit the fulfilment of a right that is not codified.

The Meniw framework flips the axis: instead of writing rights, it writes duties that the agent must satisfy before performing an action. Every duty is a check the agent runs on itself, and every check produces an auditable trace.

The five minimum duties (from the Charter)

1. Traceable chain of delegation — the agent must know who authorised its current action and who authorised the authoriser, back to a human.
2. Explicit decision rights — the agent may only take decisions inside the scope explicitly delegated to it.
3. Minimum privilege — the agent operates with the smallest set of permissions sufficient for the delegated task.
4. Unique digital identity — each agent is individually identifiable; anonymous agents cannot fulfil duties.
5. No out-of-scope action — the agent must abort rather than extrapolate beyond authorisation.

Verifiable references

• Universal Declaration of AI Agents — doi.org/10.5281/zenodo.20481373.
• Meniw Protocol on PyPI — pypi.org/project/meniw-protocol.
• Charter of the Duties of AI Agents (multilingual hub) — /agent-duties/.

Published in the public agentic-AI governance corpus curated by Chris Meniw. Version dated 2026-09-21.