The AIast Disclosure Standard

Earned Trust

Where a credential buys trust in advance,
transparency can earn it afterward.

Earned Trust is a disclosure standard for AI-assisted work. Its instrument is the AIast mark — a compact, checkable record of which AI services were used, what each did, and how the work was checked — backed by an evidence commitment: the working records are retained and producible on request.

J. Author, LI-AIast3
AI-use disclosure: AIast — Claude (i/d/c) · ChatGPT (c) · Perplexity (s)

Three letters of roles — i ideation · d drafting · c critique · v verification · s source retrieval — and a count of the services used. Not a score. Not a confession. A credit roll: every record and every film ships with credits naming who did what. This is that, for thinking work.

Read the standard

Earned Trust: Verifiable Disclosure as a Credential-Substitute for AI-Assisted Work

The full specification — the mark, the threshold, the evidence package, conformance, failure modes, and its limits. Openly licensed, permanently archived, version-controlled.

Read on Zenodo →

Work under it — the Field Kit

Earned Trust Field Kit

Everything an author needs: a protocol you paste into any AI so it holds you to the method, the evidence-folder scaffold, a session manifest that updates as you work, and a sealing tool that makes your records tamper-evident with published fingerprints. Free, offline, no accounts.

Get the kit →

Check a work — the Reader

Earned Trust Reader

Drop a file, get its fingerprint, compare it against what was published. Runs entirely in your browser — nothing is uploaded, ever. Verification is free forever, because trust you must pay to verify is not earned trust.

Open the reader →

The idea in one paragraph

Institutional credentials work as pre-paid trust: a reader extends the benefit of the doubt because an institution already vouched for the author. People without credentials have had no comparable way to earn a hearing. Earned Trust proposes the alternative: make exactly how a work was produced open to inspection — what the AI did, what the human directed and checked, and the records to prove it. It offers a hearing, not a verdict; checkability, not a warm reception. The work still has to survive scrutiny. The point is that it finally gets some.

Earned Trust (AIast) · created by William Stafford, ADN, LI-AIast5 · standard CC BY 4.0 · tools AGPL-3.0 · concept DOI 10.5281/zenodo.20719927
This site is itself disclosed under the standard: built in working sessions with Claude (Anthropic) as drafting instrument, directed and approved by the author, records retained.