Skip to content

FAIR² Specification

Welcome to the official documentation for FAIR² (FAIR Squared) — the metadata specification that advances the FAIR principles for datasets requiring context-rich, AI-ready, and responsible reuse.

FAIR² makes datasets not only Findable, Accessible, Interoperable, and Reusable, but also:

  • Machine-Actionable — compatible with automated workflows
  • Provenance-Aware — enriched with structured provenance and processing trails
  • Responsibly Governed — aligned with ethical, legal, and licensing requirements for AI use

What is FAIR²?

FAIR² is a community-driven specification designed to meet the needs of modern research, government, and industry data ecosystems. It:

  • Structures metadata to be useful in ML/AI workflows
  • Supports automated compliance checks, access gating, and responsible reuse
  • Includes support for computable licenses and policies using ODRL
  • Extends Schema.org and aligns with DCAT, DataCite, and ODPS

Key Features

  • Context-Rich Metadata: Capture details on data generation, cleaning, validation, and ethical limitations
  • AI-Ready Design: Optimized for ML workflows, including descriptor integration and method linking
  • Responsible AI Alignment: Support for auditability, access control, and consent via access levels and policy layers
  • Standard Compatibility: Works with ML Croissant, Schema.org, and other GO FAIR supported standards

Core Concepts

FAIR² defines:

  • Dataset Shapes: Metadata templates extending schema:Dataset with domain- and use-specific descriptors
  • Method Structures: Stepwise, provenance-aware method documentation (e.g., for computational workflows)
  • Agreement Levels: A 4-tier access control framework (Open → Secure) linked to machine-readable policies
  • Provenance Trails: Link datasets, methods, tools, and people via prov:wasGeneratedBy, prov:used, etc.

How to Use This Documentation

Browse the documentation sections:

  • 📖 Specification: Detailed field definitions and reference models
  • 🧩 Integration Guides: Interoperability with standards like Schema.org, ODPS, and ML Croissant
  • ⚖️ Access & Policies: Using agreement levels and ODRL policies
  • 👥 Community: Contribution guide, governance, and roadmap

Try It

To see examples of FAIR² metadata in use: - View annotated JSON-LD examples in the technical section - Try a FAIR² metadata template with live schema validation (coming soon)

Get Involved

FAIR² is developed by the FAIR² Alliance, a neutral and open community.

  • Join working groups
  • Propose a pilot dataset
  • License the specification for certification or education

For contributions, see Contributing.

Roadmap

Development of FAIR² is guided by both technical milestones and community feedback. For planned features and release targets, see the Roadmap.

For additional information and technical resources, please explore the full documentation via the navigation menu or the links above.