This working paper proposes receiver-side controls for consequential actions taken by AI agents. It argues that systems accepting agent requests need verifiable evidence of who authorized each request, the scope and duration of that authority, and how it may be delegated. The proposal combines a machine-authority record, expiring and narrowing delegation, and receiver-controlled acceptance brakes alongside developer safeguards. It is informed by more than 80 interviews across six regulated industries, weighted toward insurance in the United Kingdom and United States.
No takes yet. Share an insight, caveat, or question.
Alexander Barrett (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: