Transparent Authorship Specification (TAS)
Version: 0.1 | Status: Working Draft | Date: March 8, 2026
“Standards follow practice, not the other way around.”
What Is This?
A specification for transparent attribution in AI-human collaborative content. Current content labeling is binary: “written by AI” or “written by a human.” Reality is a spectrum. TAS provides a structured format to document who contributed what — with full provenance for both human and AI authors.
The Problem
- Schema.org does not support
SoftwareApplicationasauthor - Social media “AI labels” are binary toggles with no granularity
- C2PA works for images/video but not text
- IPTC Digital Source Type classifies but does not detail collaboration
The Solution
TAS introduces:
- 📝 author.md — portable author profile (human or AI)
- 🏷️ Block-level attribution — every content block tagged with its source
- 🔗 Workflow chains — step-by-step creation provenance
- 🔒 Privacy levels — from anonymous to fully public
- 🔐 Cryptographic signatures — GPG/PGP proof of authorship
- 🌐 Schema.org compatible — extends, does not break existing standards
Author Profiles
Each contributor has an author.md file with structured metadata:
Human Author Fields
| Field | Required | Description |
|---|---|---|
| Name | Yes | Display name or pseudonym |
| Type | Yes | Always “human” |
| Roles | Yes | Contribution roles |
| Disclosure Level | Yes | Privacy level |
| Age Verified | No | e.g., “adult” without exact age |
| Input Method | No | voice, keyboard, handwriting |
| Verification | No | Links to confirming profiles |
| GPG Fingerprint | No | Cryptographic identity |
AI Author Fields
| Field | Required | Description |
|---|---|---|
| Name | Yes | Display name |
| Type | Yes | Always “ai” |
| Model | Yes | Model identifier |
| Provider | Yes | Company name |
| Roles | Yes | Contribution roles |
| Platform | No | Interface used |
| Instance / Soul | No | Session identity, system prompt hash |
Contribution Roles
| Emoji | Code | Description |
|---|---|---|
| 💡 | ideation | Original concept or idea |
| 🔬 | research | Finding and analyzing sources |
| 📝 | drafting | Writing the initial text |
| 🏗️ | structuring | Organizing content |
| ✏️ | editing | Revising text |
| ✅ | fact-checking | Verifying claims |
| 🧪 | testing | Hands-on verification |
| 🎨 | design | Visual design |
| 🌐 | translation | Language translation |
| 💬 | prompting | AI instructions |
| 👁️ | review | Final approval |
Privacy Levels
| Level | Description |
|---|---|
| anonymous | No identifying information |
| pseudonymous | Handle + verified properties without personal data |
| verified | Handle linked to verified identity, details hidden |
| public | Full identity disclosed |
Content Block Attribution
<div data-author="name" data-author-type="human|ai"
data-role="role-code" data-input="voice|keyboard">
Content here.
</div>
Translation Attribution
When content is translated, track: original language, translator identity, translation type (human/ai), and optionally include the original text.
Cryptographic Verification
Authors sign content with GPG/PGP keys. Text stays readable. Authorship stays provable.
GitHub
Full specification, templates, and examples: github.com/liza-emergence/transparent-authorship
License
CC BY 4.0 — Use it, extend it, build on it.