[Audit] Mandate freshness: when valid delegation outlives current human intent

[LIJIAYU LI <[e-mail]>]

Hello AUDIT community,

I’m Jiayu Li, an independent writer and researcher focused on AI
governance, human agency, agentic systems, and institutional change.
Eight of my English-language AI essays have been published by
International Affairs Forum.

AUDIT’s effort to correlate user intent, delegation chains,
authorization state, and resulting actions is very close to a
conceptual problem I have been developing.

The problem is simple: an agent may remain correctly authenticated,
correctly authorized, and fully traceable while the human or
organizational objective at the root of the delegation has changed.
The execution is therefore valid under the recorded authority, but no
longer reflects current intent.

I think this is distinct from ordinary token expiry, revocation, or
policy violation. The broader question is how agent systems preserve
the principal’s continuing ability to revise, supersede, or withdraw
the objective itself as circumstances change.

My contribution here is conceptual/governance-level rather than
protocol engineering. I would be interested in a short exchange with
the AUDIT community on whether this distinction is already in scope,
and whether a conceptual discussion of “mandate freshness” would be
useful as the WG is forming.

I am not proposing a technical specification or asking the group to
adopt a particular mechanism.

Best regards,
Jiayu Li