News · 10 September 2026
Agent-access receipts published in Frontiers in Computer Science
“Agent-access receipts for artificial intelligence-mediated content and application programming interface access: a reproducible synthetic security experiment” is published in Frontiers in Computer Science, volume 8, open access under the Creative Commons Attribution licence, doi:10.3389/fcomp.2026.1900546.
When an AI agent fetches a page or calls an API on someone's behalf, the publisher's edge usually treats it as a bot-management problem: identify the automated caller, then allow, block, delay, challenge or meter it. The article argues that the framing is useful but incomplete. A disputed access decision can turn on the caller's authenticated identity, the principal it represents, the authority delegated to it, the declared purpose, the resource or API surface, the policy snapshot in force, the sensitivity of the response, the payment or licensing state, and whether the material needed to validate all of that still exists once the request is over.
An agent-access receipt is a compact decision record generated at the edge and preserved for later inspection. The study implements a deterministic Python laboratory with a synthetic policy and a synthetic trust registry, and evaluates 15 scenario cases, 15 mutation and control cases and 3 post-issue receipt-integrity controls. In the deposited package every case matches its expected outcome and all 30 generated receipts pass the schema guard. The artefact is not a production security protocol; it claims no legal compliance, deployment readiness or third-party endorsement. What it shows is why access-control evidence for AI-mediated callers should preserve the bindings among identity, delegation, purpose, inventory and economic authorisation, rather than a bare allow-or-block verdict.