AI Explainers:
your questions answered
Other countries are racing ahead with clear AI regulations, giving businesses certainty and attracting investment. Japan already offers broad exceptions for data analysis - providing legal clarity and strong support for AI startups.1 The US permits transformative use under the fair use doctrine. Singapore allows copyrighted works to be used for computational data analysis provided the material is lawfully accessed - explicitly supporting AI growth.2 The EU has moved faster to provide clarity for developers and creators through its opt-out framework.
Meanwhile, the UK remains in limbo as the government continues to delay a decision on AI and copyright reform. This delay has consequences. Every major AI frontier model is currently built abroad. Stability AI exemplifies the challenge. Founded in Britain, it has trained models in jurisdictions such as the US and Japan where copyright and training rules are clearer. The UK’s ambiguous copyright framework creates legal risk around AI training. Where regulation is unclear, companies move to countries where the rules are defined.
The UK’s current TDM exemption applies to non-commercial research where the copy is accompanied by a sufficient acknowledgement. That makes it incompatible with how modern AI is built and deployed. This exemption allows universities and other researchers to analyse lawfully accessed copyrighted works for research. But commercial innovators - including startups in areas like health, science and public services - operate in legal grey areas. This means promising AI systems risk remaining stuck in research environments rather than benefiting patients, students and citizens in real-world services.
Legal clarity would particularly benefit UK startups, researchers and small firms which don’t have the resources to navigate this uncertainty. Uncertainty favours only those who can afford prolonged litigation.
The Getty Images v Stability AI case, ruled on by the High Court in November 2025, examined how copyright laws affect the development of generative AI models.3 Getty claimed that its images were used to train a generative AI model developed by Stability AI - and that the AI could reproduce Getty’s trademarks. However, the training of the model did not take place in the UK, and as such the court didn’t pass judgement on whether training AI models in the UK infringes copyright. The case is a clear warning sign. Where the rules aren’t clear or competitive, AI companies will simply train their models overseas - so these models continue to be developed, but outside the UK. Without reform, more British AI companies will train and scale abroad - rather than allowing the UK to embrace and profit from them at home.
Implementing a mandatory licensing regime for AI would be complex, costly and unworkable at scale. AI models are trained on vast and diverse datasets, often comprising millions or even billions of datapoints. Creating and enforcing a licensing system would require constant updates and impose high compliance costs - disproportionately favouring the large AI developers who can absorb them and penalising innovative British startups. It would limit access to data, lower the quality of models, and undermine the work of UK developers - 94% of whom rely on models using TDM from publicly available data.4 It would slow innovation without delivering meaningful protection for creators whose legally accessible data could still be used to train AI models abroad. Moreover, effective AI licensing markets can only emerge if significant AI training activity actually takes place in the UK. Simply mandating a licensing regime will not create that activity.
The creative and technology sectors are deeply intertwined. AI is already a core part of how the creative industries grow. Just as user-led streaming services, such as YouTube, produced new creative careers and global audiences, AI is expanding creative output, reaching new audiences, removing barriers to entry, and enabling new forms of cultural innovation. Around 40% (£49.1bn) of the creative sector’s £124 billion value to the UK economy (GVA) comes from IT, software and computer services. The same sub-sector accounts for 42.6% of creative industry jobs, with wages around 50% higher than the average UK wage.5
AI development is global. UK policy cannot stop AI training - it only determines where it is developed, and who benefits. Stricter copyright laws in the UK would not prevent legally accessible data from UK creators being used to train AI models abroad that are then made available in the UK. The result is simply fewer jobs and less investment in the UK. A clear, pro innovation framework would keep development in the UK, ensuring creators benefit from new technologies and new audiences.
AI models do not contain libraries of copyrighted works. During training, they learn patterns from legally accessible data, similar to how humans learn language. Original works are not retained inside the model, and copyright law continues to apply to outputs and commercial uses. A workable solution anchored in a TDM exemption does not remove privacy or data protection obligations. It simply provides legal clarity and enables innovation.
Appendix
1 Centre for British Progress, Copyright & AI: The Case for a Pro-Growth Approach, February 2025
2 Tech Policy Press, AI Training and Copyright Infringement: Solutions from Asia, October 2024
3 Mayer Brown, Getty Images v Stability AI: What The High Court’s Decision Means For Rights-
Holders And AI Developers, November 2025
4 Computer & Communications Industry Association, UK AI Ecosystem Poll Shows The Importance Of
Copyright And AI Regulation For The Government’s Objectives Of Promoting UK Innovation And
Economic Growth, May 2025
5 House of Lords Library, Creative industries: Growth, jobs and productivity, January 2025