Why Transparency Matters
AI tools are transforming how we create software, data, documents, and media. A clear, standard way to declare how they were used benefits everyone.
Future-Ready Compliance
Transparency requirements are taking shape across jurisdictions. The EU AI Act, the US NIST framework, and UNESCO's AI Ethics Recommendation all point in the same direction. A structured, machine-readable declaration keeps you ahead of the curve.
Reproducible by Design
AI involvement changes how work is produced. Declaring it clearly strengthens your methodology and makes your results easier to reproduce and review.
Build Trust with Your Community
Contributors, users, and reviewers appreciate knowing how a project was built. Transparent AI disclosure strengthens collaboration and credibility.
One Format for Every Project
Instead of ad-hoc README mentions or guesswork, a single standard format makes AI disclosure consistent, complete, and easy to verify.
Built for Real Workflows
Machine-Readable
Your CI pipeline validates the declaration on every push, the same way it runs tests. JSON Schema Draft 2020-12.
Dual Format
Write in YAML with comments for your team. Export JSON for your tooling. Same schema validates both.
Linked Data
Every field in the schema maps to W3C PROV-O, Schema.org, SPDX, or Dublin Core via a JSON-LD context. Auditors and compliance tools can query your declarations with standard RDF tooling without any custom parsing.
Policy Aligned
Structured fields for EU AI Act Article 50, NIST AI RMF, and UK Pro-Innovation. Fill in the compliance section and your DPIA reviewer has what they need.
Granular
Declare per tool, per component, per scope. Not just 'we used AI' but which tool, for what, and how much.
Extensible
Covers 'no AI used' through full enterprise compliance. Works for software, datasets, documents, media.
Examples
Start from a battle-tested template. Standardized examples for every project type and complexity level.
Minimal CLI / No AI
For simple software tools that do not utilize any AI components. Demonstrates transparency even when AI is absent.
Web App
Standard template for modern web applications covering common integration points for AI assistance.
Research Project
Tailored for academic or experimental projects where data source and methodology transparency are critical.
High-Risk AI
Systems with significant societal impact, requiring rigorous disclosure of safety and ethical considerations.
Enterprise / Late-Stage
Comprehensive declarations for mature products requiring detailed auditing and compliance data.
Comprehensive
A full-feature template utilizing all available fields in the schema for maximum transparency.
How AI Declaration Format Compares
AI Declaration Format is complementary to existing standards, not a replacement.
| Standard | Focus | Relationship to AI Declaration Format |
|---|---|---|
| codemeta.json | Software metadata | Complementary. aidecl adds AI-specific disclosure to project metadata. |
| CITATION.cff | Academic citation | Complementary. aidecl covers the development process, CFF covers attribution. |
| SPDX 3.0 AI Profile | AI components in software | SPDX declares AI as a component; aidecl declares AI used to build the work. |
| CycloneDX ML-BOM | ML model inventory | ML-BOM inventories AI models; aidecl documents how AI tools were used during creation. |
| C2PA | Content provenance | C2PA tracks media provenance; aidecl tracks development and creation provenance. |
| W3C PROV | General provenance | aidecl uses PROV-compatible semantics via JSON-LD context. |
Comparison based on published documentation and specifications as of March 2026.
The AI Declaration Format Ecosystem
AI Declaration Format Schema
The specification. JSON Schema + JSON-LD context + 10 example files. The foundation everything else builds on.
aidecl CLI
pip install aidecl. Create, validate, and convert AI Declaration files from the command line. Integrates with CI/CD and pre-commit hooks.
Get Started in 2 Minutes
1. Install the CLI
$ pip install git+https://github.com/ai-declaration/cli.git2. Create a declaration
$ aidecl init --format yaml
Created aidecl.yamlHere is what it looks like:
schema_version: "1.0.0"
project:
name: my-project
content_type: software
ai_usage:
used: true
summary: "Used for code completion"
level: minimal
tools:
- name: GitHub Copilot
type: assistant
declaration:
date: "2025-03-15"
declared_by: Dev Team3. Validate
$ aidecl validate aidecl.yaml
PASS [yaml]: aidecl.yamlShow Your Adoption
Add this badge to your README to show your project uses AI Declaration Format.
View top-level schema fields
| Field | Type | Description |
|---|---|---|
schema_version | string | Version of the AI Declaration schema |
project | object | Project metadata: name, version, repository URL |
ai_usage | object | AI usage summary and list of tools used |
declaration | object | Who made this declaration, when, and review info |
compliance | object | Regulatory compliance flags (EU AI Act, NIST) |
security | object | Security review and data handling information |
Frequently Asked Questions
Is this required by regulations?
Not yet for most cases, but regulations like the EU AI Act are moving in this direction. Having a standard format in place now means you are ready when requirements formalize.
How precise does it need to be?
The declaration is a living document. Update it as your understanding evolves. Honest, approximate declarations are perfectly fine and far more valuable than none at all.
Do I need to update it on every commit?
No. Update when AI usage changes significantly: when you add or remove an AI tool, change how you use it, or modify your workflow substantially.
Is it compatible with codemeta.json?
Yes, they are complementary. codemeta.json covers general project metadata; aidecl covers AI usage specifically. Both can live in the same repository.
Who reads this file?
Auditors checking compliance, collaborators understanding the project, users assessing trustworthiness, CI pipelines running automated checks, and compliance tools generating reports.
Why YAML?
YAML is human-friendly and supports comments, making declarations easier to write and review. JSON is also fully supported with the same schema. Use whichever your team prefers.
What about SBOM?
AI Declaration Format is complementary to Software Bill of Materials. SBOM inventories software components; aidecl declares how AI was used during the creation process. They answer different questions.
Show How Your Work Gets Made
Whether you are a solo developer, a research team, or an enterprise, AI Declaration Format gives you a simple, standardized way to document AI tool usage in your work.
- Developers: Add an aidecl.yaml to your next project. It takes 2 minutes.
- Research teams: Meet reproducibility and compliance requirements with machine-readable declarations.
- Organizations: Integrate AI Declaration Format into your CI/CD pipeline for automated transparency checks.
Join the AI Declaration Format community
AI Declaration Format is an open project. Contributions, feedback, and sponsorship are welcome.
Who Is This For?
AI Declaration Format works for anyone who creates digital work and wants to be transparent about AI involvement.
Researcher
You publish a dataset that was partially cleaned with LLM assistance. An aidecl.yaml in the repository makes your methodology clear and your work more credible, answering reviewer questions before they need to ask.
Developer
Your open source project uses Copilot for boilerplate and test generation. The declaration file gives contributors confidence in the codebase by showing exactly which parts had AI involvement.
Team Lead
Instead of spreadsheets and manual audits, each repository carries its own machine-readable disclosure. Your team is ready for EU AI Act transparency requirements with minimal extra effort.
AI Declaration Format is a new standard. Be among the first to adopt it. Add an aidecl.yaml to your project and let us know.
Early adopters will be featured on this page with their project name and link.
Become an Adopter