Skip to content
All Workstreams
Workstream

Licensing & Policy

Developing modular frameworks that make permissions explicit, portable, and operational across jurisdictions.

The problem

Silence is not a license.

The current legal environment leaves AI training in a grey zone. Most training data was collected without explicit AI-use permissions because those permissions didn't exist, or weren't sought, or were buried in terms of service that nobody reads.

Courts in the US, EU, and UK are actively testing the boundaries of fair use, authorship, and liability as they apply to AI training. The outcomes are inconsistent, jurisdiction-dependent, and slow. Institutions cannot wait for legal clarity that may not arrive for years.

"The derived-work test, the fair use doctrine, the idea-expression dichotomy none of these frameworks were designed for systems that ingest millions of works simultaneously and produce outputs that approximate the aggregate. We are in genuinely new legal territory."

What's missing is not more litigation. It's a licensing layer that makes permissions explicit, portable, and operable before the question of legality even arises.

The AMPL framework

Modular tools for explicit permissions.

AI Commons is developing the AMPL (AI Model and Data Permission Layer) framework: a set of modular contract tools that make permissions explicit, portable, and operational across legal traditions.

AMPL is not a single license. It is a system of composable elements that institutions, platforms, and creators can combine to express exactly what they permit, for what purposes, under what conditions, with what compensation or attribution requirements.

  • Base license the foundational permission structure covering training, inference, and distribution
  • Modular extensions sector-specific or use-case-specific additions (commercial use, public interest, research, etc.)
  • Jurisdiction annexes adaptations for civil law systems, EU frameworks, and non-Anglo-American IP traditions
  • Machine-readable encoding SPDX-compatible, schema.org-integrated, platform-implementable
  • Human-readable layer plain language summaries that creators and communities can actually understand

AMPL is the maturation of a licensing mechanism AI Commons first attempted in 2018. The legal environment then was not adequate to the ambition. The combination of the EU AI Act, active litigation, and institutional demand has made this moment the right one.

Policy engagement

Standards that governments can require.

The licensing framework is not only a tool for voluntary adoption. AI Commons engages directly with policymakers to ensure that provenance and licensing standards can be written into procurement requirements, regulatory frameworks, and institutional governance policies.

The EU AI Act's enforcement in August 2026 creates the first major compliance forcing function. AI Commons is working to ensure that AMPL is ready to serve as operational infrastructure for that compliance, not just aspirational guidance.