As AI systems begin to make decisions, enter agreements, initiate transactions and manage workflows on behalf of a person or organisation, a new question follows: who authorised the machine to act, and does that authority still hold? Machine Authority is the concept that captures this — the standing of a machine actor to rely on a trust signal, or to act, on behalf of whoever authorised it.
What has to be demonstrable
For a machine actor, the same questions apply that would apply to a person or organisation exercising authority: who authorised the system, what authority was delegated to it, what limitations apply, whether that authority remains valid, what decisions were made under it, and what evidence supports those decisions. Historically these were implicitly answered by a human being in the loop. When the actor is a machine, they need to be explicit and checkable.
Why this doesn't require new principles
Machine Authority does not require a separate governance model from the one that already governs people and organisations — it applies the same requirement (authority must be established, scoped and provable before it is exercised) to a new category of actor. See Delegated Authority for how that authority actually gets handed to a machine in the first place.