Answers are grounded in this paper and cite their sources.
This draft specification states the rules a server follows to refuse an AI agent's write unless an approval exists for that exact change. The server hashes the change, refuses it without a matching approval, and lets a person approve it on the server's own page or through a signed receipt, with single use, bounded grants and a hash-chained audit record. The release includes conformance tests, test vectors and a sandbox server. Draft 0.1. The specification text is licensed under CC BY 4.0; the code under sandbox/ and vectors/ is licensed under the Apache License 2.0.
No takes yet. Share an insight, caveat, or question.
Sharoze Iftikhar (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: