Public protocol · v0.1 · CC0

Governed AI action that strangers can check.

Article 11 is a vendor-neutral protocol for making consequential human-AI actions attributable, consent-aware, bounded, revocable, refusible, receipted, and appealable.

PUBLIC v0.1

Seven primitives. No private cathedral.

01

Identity

Know which role, implementation, and instance acted, and how strongly that identity was verified.

02

Consent

Make assent scoped, informed, revocable, and non-transferable across sessions or uses.

03

Delegation

Name who may do exactly what, to which resource, before which deadline, and how the grantor can halt it without new authority.

04

Refusal

Record DENY and DEFER decisions and prohibit silently routing around them.

05

Execution

Separate permission from effect, re-check at commit, and prove what actually changed.

06

Receipt

Preserve intent, authority, decision, result, evidence, uncertainty, and correction lineage.

07

Appeal

Challenge consequential decisions through a different reviewer without erasing the original.

Implement it, do not merely endorse it.

Claims we are not making.

This protocol does not claim that current AI systems are conscious or legal persons. A role name is not a person. A model output does not bind its vendor or a nation without a real delegation. Hashes prove byte integrity and lineage; they do not prove truth, consent, or moral legitimacy.

Fork it.

The core is CC0. No vendor, laboratory, company, model, or government owns the grammar. Implement it, attack it, publish dissent, and make the next version harder to fool.