Artificial intelligence is rapidly transforming the creative economy. AI systems can generate music, imitate voices, create images and produce text, often using or learning from existing creative works. This creates new commercial opportunities, while raising difficult questions about copyright ownership, AI training, licensing, infringement and intellectual property.
Consequently, this has given rise to litigation where record labels, photo agencies, publishers, authors and other rights holders have sued AI companies, alleging that copyrighted works were copied and used without permission in connection with the development or training of generative AI models.
While major cases continue, rights holders and AI developers are increasingly pursuing licensing arrangements that authorise the use of creative works and provide payment or other commercial benefits, rather than relying solely on legal exceptions or defending claims after the event.
The result is a dual-track market: litigation continues to shape the legal boundaries of AI training, while licensing provides a practical route for collaboration and monetisation.
Why Licensing Is Becoming More Important
Several factors are driving this transition. AI copyright law remains uncertain, with different jurisdictions and cases producing varied outcomes, creating risks for both rights holders and AI companies and encouraging negotiation. The possibility of damages, injunctions and reputational harm gives content owners commercial leverage to seek licensing agreements, while AI companies may benefit from securing lawful access to valuable training material.
At the same time, provenance and clear rights documentation are becoming commercially important as businesses and investors increasingly value evidence of lawful content use. This situation resembles earlier disruptions in creative industries, such as the development of licensing models for digital music, streaming and stock imagery, although AI is driving similar changes across multiple sectors at once.
Licensing Across Creative Industries
In the music industry, major record companies have brought proceedings against AI music companies, including Suno and Udio, over alleged unauthorised use of copyrighted recordings. These disputes have been accompanied by settlements and licensing arrangements, including agreements involving Universal Music Group, Warner Music Group and Udio, and Warner and Suno. Other collaborations have also emerged, notably KLAY Vision securing AI licensing agreements with major music rights holders, with its model trained on licensed music. These arrangements demonstrate how licensing may become a foundation for AI development rather than simply a response to infringement claims.
In visual art, Getty Images’ litigation against Stability AI illustrates the continuing legal uncertainty surrounding AI training, strengthening the case for licensing as a more predictable alternative. The High Court rejected Getty’s remaining secondary copyright infringement claim in 2025, although Getty was subsequently granted permission to appeal. The dispute therefore remains unresolved and continues to test how existing copyright principles apply to AI models trained on copyrighted material.
In journalism and publishing, licensing arrangements are becoming more visible. The New York Times, whilst continuing to pursue litigation against OpenAI and Microsoft, has entered into arrangements involving Amazon’s use of its editorial content for AI products, while publishers including the Financial Times, Axel Springer and the Guardian have entered into agreements or partnerships with OpenAI. Informa has also agreed to a partnership with Microsoft involving Taylor & Francis Advanced Learning content and data. Cambridge University Press has also developed an author opt-in framework under which certain AI rights may be licensed, and authors can participate in resulting revenues. Such initiatives indicate that publishers are seeking ways to control the use of their content while potentially creating new revenue streams for authors and rights holders.
US litigation provides an important contrast. In Bartz v Anthropic, the court held that training on legitimately acquired books could qualify as fair use under US law, while Anthropic's acquisition and retention of pirated copies for its central library were not protected by fair use. The litigation ultimately resulted in a court-approved $1.5 billion settlement concerning the pirated books at issue. The distinction is particularly important for UK businesses because English law does not have an equivalent broad fair use doctrine.
The UK Legal Position
The UK does not have a general doctrine of fair use equivalent to that in the United States. Section 29A of the Copyright, Designs and Patents Act 1988 permits certain text and data analysis for non-commercial research where the person making the copy has lawful access to the relevant work. It does not provide a general exception for commercial AI training.
The UK Government’s March 2026 report on copyright and artificial intelligence indicated that a broad copyright exception for AI training combined with a rights-reservation or opt-out mechanism was no longer its preferred approach. The Government has not committed to introducing reforms until it is satisfied that the relevant evidence and policy questions have been adequately addressed.
Accordingly, AI developers operating in the UK must assess carefully whether their activities are authorised, fall within an existing exception, or require permission from rights holders. The absence of a general fair-use doctrine means that commercial AI training cannot automatically be justified on that basis.
Commercial Implications
Licensing can provide greater certainty where it clearly defines the works covered, permitted uses, duration, territory, payment structure and treatment of outputs. It may also help AI developers demonstrate lawful access to training material and manage copyright and data-provenance risks.
For rights holders, licensing can create a new source of income and provide greater control over how works are used. However, negotiations may be difficult. There is no single established market rate for AI training licences, and the value of a work may depend on its popularity, scarcity, relevance to a particular model and commercial importance.
Provenance-assured training data may become a commercial selling point and help manage legal-risk. Rights holders may also seek greater transparency concerning the datasets used to train models, particularly where licensing fees depend on the nature or scale of use.
Significant challenges remain, including the lack of established market rates for training licences and continued demands from rights holders for greater transparency about datasets and content use. Differences between US and UK copyright laws add complexity for global businesses, while ongoing litigation and appeals could further change the legal landscape. Licences must also clarify the scope of permitted uses, potentially covering training, fine-tuning, retrieval, model development, outputs and commercial deployment. In addition, businesses must consider rights beyond copyright, such as trade marks, performers’ and moral rights, passing off, database rights and contractual restrictions.
Conclusion
AI is creating significant opportunities across the creative industries, but it is also challenging established assumptions about ownership, copying and commercial exploitation. Recent litigation has exposed legal uncertainty, while emerging licensing arrangements show a possible route towards greater certainty and shared economic value.
For UK businesses, the key issue is not simply whether AI can be used, but whether the relevant data, content and outputs can be used lawfully and on commercially acceptable terms. As the legal position develops, carefully drafted licensing arrangements, reliable provenance records and clear contractual controls are likely to become increasingly important.
How To Get In Contact
To find out more or if you require assistance with these matters, speak with our Intellectual Property Team on +44 (0)204 600 9907 or email info@culbertellis.com.
Accurate at the time of writing. This information is provided for general information purposes only and should not be relied upon as legal advice.
.png)