What does human authority over AI mean?
Human authority over AI means that the legitimate right to cause an action originates with accountable humans and organisations and is delegated to AI systems only in scoped, revocable, auditable form. AI has no authority by default; capability, identity and autonomy never create authority on their own; and high-impact actions require policy-controlled human authorization.
- AI has no authority by default.
- Capability does not equal permission.
- Identity does not equal authorization.
- Autonomy does not create authority.
Frequently asked questions
Is human authority the same as human-in-the-loop?
Not exactly. Human-in-the-loop means a person reviews actions. Human authority means all authority originates with people and is delegated by policy; approval is required where consequence demands it, not for every action.
Does human authority slow AI down?
Low-risk actions within granted capabilities execute immediately. Only high-impact actions are held for approval.