The bit passport

Proposal · not yet in the parser

View as Markdown Open in Claude ↗

The bit passport is one header at the top of a delivery that applies to every bit in it: who holds the rights, under which licence, what AI may do with the content, and which accessibility standard it meets. What describes a single bit's content stays in that bit.

What it carries

In the passport, once per delivery Stays in each bit
Rights: copyright, licence, rights policy, publisher Identity: the bit's ID and the customer's ID
AI use: what agents may do (retrieve, quote, cite, train) The AI label and the machine-translation label a reader must show
How the content was made: AI model, operation, human review The language of a passage
Accessibility claim: conformance, certifier, features, hazards, summary Accessibility evidence: alt text, decorative marks, table headers
Origin: source document, edition, translation of Where the bit sits in the original
Audience, when it holds for every bit: language, age, level

A bit never overrides the passport's rights: narrowing the rights for a single bit makes no sense. Content under other rights goes into a delivery of its own, with its own passport.

Shape on the wire

The passport is the first line of a JSONL delivery, under its own key, so a reader tells it apart from the bits that follow. The names are drafts:

json
{"passport": {"@context": ["https://schema.org"], "publisher": "Example Publisher", "license": "https://creativecommons.org/licenses/by/4.0/", "inLanguage": "en", "conformsTo": "EPUB Accessibility 1.1 - WCAG 2.2 Level AA", "aiUse": {"retrieve": true, "train": false}}}
{"bit": {"type": "p", "id": ["a1"], "body": "…"}}
{"bit": {"type": "p", "id": ["a2"], "aiGenerated": true, "aiDisclosureRequired": true, "body": "…"}}

The passport is planned as JSON-LD, so it can reuse vocabularies that exist already instead of new names: schema.org for publisher, licence, language and accessibility, and the W3C's policy and text-and-data-mining vocabularies for what AI may do. Compact JSON-LD reads as plain JSON, so no linked-data tools are needed. The bit lines stay plain JSON.

The accessibility claim

The passport states the claim; the bits carry the evidence, so the claim can be checked. It uses the terms EPUB Accessibility already defines, so an EPUB export takes them over unchanged:

  • conformance, for example "EPUB Accessibility 1.1 – WCAG 2.2 Level AA";
  • who certified it, with their credential and report;
  • the access modes and which of them suffice alone;
  • the features present: alternative text, long descriptions, transcripts, captions, table of contents, reading order, structural navigation;
  • hazards: none, flashing, motion or sound;
  • a plain-language summary.

Today

Until the passport ships, these properties are bit tags: @copyright, @license, @rightsPolicy, @publisher, the AI tags such as @aiGenerated and @aiModel, and the accessibility and audience tags. The reference lists them on every bit page. The working groups of the bitmark Association each own one section of the passport.