The UK's suggested “Commercial Research Exception” (CRE) for AI training is not a workable middle ground. It either (a) blocks most commercial releases due to licensing holdouts...

2025-11-28 · Source: Pascal’s Substack · Field: Legal & Regulatory — Intellectual Property & Patents, Regulatory Affairs & Government Relations · Depth: Advanced, medium

Summary

A UK "Commercial Research Exception" (CRE) for AI training, which proposes allowing AI developers to train on copyrighted works without upfront permission but requiring licensing at "market entry," is argued to be unworkable. This approach, while seemingly a pragmatic compromise to foster innovation, is predicted to either block most commercial AI releases due to rights holder refusals or inevitably transform into politically contentious compulsory licensing. The core issues stem from the capital-intensive, irreversible nature of AI training, which shifts bargaining power and auditability downstream. The concept of "market entry" is also problematic, as modern AI productization involves porous boundaries like synthetic data generation, internal models powering services, and non-public commercialization, creating numerous loopholes for evading licensing obligations.

Key takeaway

For AI/ML directors evaluating regulatory landscapes, understand that "license later" frameworks for AI training introduce significant investment risk and regulatory uncertainty. Your teams could spend millions on training only to find models legally unreleasable or subject to unpredictable licensing demands. Prioritize engaging with frameworks that support upfront licensing, clear provenance, and public investment in lawful data access to ensure long-term stability and avoid potential legal and reputational backlash.

Key insights

A "license later" approach for AI training is unworkable, leading to either blocked releases or forced licensing.

Principles

• AI training is capital-intensive and path-dependent.
• Unanimity among rights holders is unrealistic at scale.
• Competitiveness policy should not weaken IP rights.

In practice

• Support scalable licensing infrastructure for AI training.
• Fund public-interest datasets with clear rights.
• Strengthen provenance expectations for AI models.

Topics

• AI Copyright Law
• Commercial Research Exception
• AI Model Licensing
• Intellectual Property Rights
• AI Policy & Regulation

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Policy Maker, Legal Professional, AI Product Manager

Related on AIssential

• Preserving the Freedom to Learn in AI — Preserving the "freedom to learn" is crucial for an open and competitive AI ecosystem.
• The most profound and least understood risk facing AI developers is endogenous: the very technology they are building is rapidly acquiring the computational and analytical capability... — AI's inherent analytical power is exposing its creators' unsustainable financial, legal, and technical vulnerabilities.
• California’s AB 412 Still Demands Developers Do The Impossible — California's A.B. 412 bill imposes an unworkable copyright disclosure burden on AI developers, hindering innovation and favoring large tech.
• BMG sues Anthropic for using Bruno Mars, Rolling Stones lyrics in AI training — AI training on copyrighted content faces increasing legal challenges, particularly from aggressive copyright holders like music labels.
• Australia’s official plan for AI safety isn’t much more than a single dot point. Will it be enough? — Australia's AI regulation relies on existing laws, contrasting with other nations' specific AI legislation.
• The Guardian view on changes to copyright laws: authors should be protected over big tech | Editorial — Authors are demanding robust copyright legislation to protect their work and livelihoods from unlicensed AI use.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Pascal’s Substack.