Introduction
The UK’s Copyright Licensing Agency (CLA) is set to launch a groundbreaking Generative AI Training Licence in the third quarter of 2025. This initiative aims to address the challenges of training large language models (LLMs) by balancing the interests of content creators and AI developers [4], ensuring fair remuneration for the use of copyrighted works while maintaining copyright protection.
Description
The UK’s Copyright Licensing Agency (CLA) is developing a pioneering Generative AI Training Licence [4] [5] [11], the first of its kind in the UK [2] [3], set to be released in the third quarter of 2025 [5]. This initiative aims to address the challenges of training large language models (LLMs) by balancing the interests of creators with the needs of developers [4], ensuring that publishers and authors receive fair remuneration for the use of their works while maintaining copyright protection. The licence is designed to provide a scalable collective licensing solution [4], offering legal certainty for AI developers to utilize a wide range of content for innovating and training their models [5]. It will cover activities such as AI model training [8], fine-tuning [2] [6] [7] [8] [10], and retrieval-augmented generation (RAG) [2] [8] [10], establishing a clear legal pathway for the use of copyrighted works in generative AI applications [8].
In collaboration with the Publishers’ Licensing Services (PLS) and the Authors’ Licensing and Collecting Society (ALCS) [4] [5], the CLA seeks to maximize the value of published content through collective licensing [4]. PLS represents 4,500 publishers and focuses on enhancing the value of their works [4], while ALCS collects royalties for secondary uses of writers’ work on behalf of its 125,000 members [4]. This collective licensing framework represents a significant advancement in the intersection of AI and copyright law [11], offering a structured legal pathway for AI developers to access copyrighted material while ensuring compensation for rightsholders who may lack the means for direct negotiations with technology firms.
A key feature of the licence is the “retrospective rights option,” which enables developers to regularize previously unauthorized uses of copyrighted material [11], thereby addressing past grievances and improving relationships between content creators and technology developers [11]. This approach is particularly relevant in light of ongoing legal disputes [11], such as those involving Getty Images and Stability AI [11], which highlight the friction between technological innovation and existing copyright frameworks [11].
Rights holders will need to register their works to participate in this collective licensing model, which is designed to complement existing bespoke licensing agreements [6] [7]. This initiative is expected to particularly benefit smaller creators and independent publishers who may struggle with individual negotiations, empowering them by providing access to language training resources and addressing their limited bargaining power [4]. Compensation models are currently under negotiation [6], aiming to strike a balance between affordability for AI developers and fair remuneration for creators, with fees distributed among ALCS and PLS members based on established rules [7].
Additionally, a national rights reservation registry is being developed to support machine-readable licensing metadata and signal licensing preferences [6] [7]. Licensees will be required to adhere to transparency standards [6] [7], including reporting on the works used and methods of content acquisition [7], fostering trust between the creative industries and AI developers [6] [7]. The CLA is also expanding its suite of licences for commercial text and data mining and workplace use of content in generative AI prompts [8], with the first of these launching on May 1, 2025 [8], reflecting a concerted effort within the industry to protect content creators as AI technology evolves [9].
While primarily focused on the UK, the licensing framework is designed with international interoperability in mind [6], potentially serving as a model for future cross-border AI licensing solutions [6] [7]. This initiative aims to reconcile the needs of AI innovation with copyright law [6], offering a voluntary and transparent alternative to statutory exceptions while safeguarding creators’ economic interests and enabling responsible AI development [6]. The success of this initiative will depend on the pricing models established for the licences and their affordability for smaller developers [11], as well as the adequacy of compensation for creators and the biases inherent in the data used for AI training [11]. Overall, the Generative AI Training Licence represents a proactive approach to integrating legal and technological advancements [11], potentially reshaping the dynamics between tech companies and content creators and influencing global copyright policies.
However, the proposed collective licensing framework has sparked controversy, particularly regarding the opt-out system that would allow works to be used in AI training by default unless authors explicitly request exclusion [1]. Critics argue that this approach undermines a core principle of intellectual property [1], which asserts that copyright protection is automatic upon creation [1], and that any objection to reproduction should be enforceable without requiring prior consent for protection [1].
Conclusion
The Generative AI Training Licence represents a significant step forward in harmonizing the interests of AI developers and content creators. By providing a structured legal framework [11], it aims to facilitate innovation while ensuring fair compensation for creators. However, the initiative’s success will depend on its ability to address concerns about automatic inclusion in the licensing system and the balance between affordability for developers and adequate remuneration for rights holders. As the framework evolves, it may serve as a model for international AI licensing solutions, influencing global copyright policies and the relationship between technology and intellectual property.
References
[1] https://olartemoure.com/en/collective-license-for-ai-training/
[2] https://librarylearningspace.com/united-kingdom-publishers-licensing-services-and-authors-licensing-and-collecting-society-create-pioneering-ai-licence/
[3] https://www.stm-publishing.com/pls-and-alcs-agree-to-development-of-pioneering-cla-generative-ai-licence/
[4] https://www.iplawwatch.com/2025/04/28/could-this-be-the-ai-nswer-a-collective-copyright-licence-for-generative-ai-training/
[5] https://www.publishingscotland.org/2025/05/cla-pls-and-alcs-to-introduce-new-license-for-genai-training/
[6] https://natlawreview.com/article/uks-collective-licensing-initiative-aims-harmonize-ai-and-copyright-law
[7] https://www.jdsupra.com/legalnews/uk-s-collective-licensing-initiative-2987670/
[8] https://downthetubes.net/publishers-licensing-services-authors-licensing-and-collecting-society-partner-with-copyright-licensing-agency-on-generative-ai-licence/
[9] https://tribune.com.pk/story/2542496/with-a-collective-licence-uk-authors-get-a-slice-of-the-ai-pie
[10] https://www.linkedin.com/pulse/pls-alcs-support-development-uks-first-collective-licence-straive-9wjvf
[11] https://opentools.ai/news/uks-cla-rolls-out-ai-training-license-a-new-chapter-for-copyright
