AI and copyright law face uncertain future

The government has recently stepped back from its previous proposals to “support growth and improve living standards, while protecting human creativity and our world-leading creative industries.” The position on the use of copyright works for training AI models remains unresolved, however industry practices and rapidly evolving competing AI economies are advancing at such a rate that legislative change in some form may be inevitable. 

This article takes a look at the legal framework and business practices as matters stand, the ongoing policy debate, and why professionals should continue to look at this carefully in order to protect their respective positions and clients’ interests.

Background to copyright

Copyright is a constantly evolving area of law. Since its roots the 1710 Statute of Anne which gave authors a statutory right to control who could reproduce their book, the law relating to copyright has adapted as technology has given rise to new types of media, business practices, and ultimately ways to copy and communicate creative works. The emergence of artificial intelligence presents another such challenge, as AI tools become increasingly capable, accessible, and widely deployed. 

The topic of generative AI and copyright is complex but increasingly relevant. New possibilities are opened for legal service provision, and for existing practitioners to reduce costs and meet evolving client demands. This article is therefore limited to the licensing practices and their interaction with AI model training, and potential reform to the law protecting computer-generated works. 

Copyright in legal documents produced with assistance of AI may be affected which practitioners will need to be aware of. Similarly, copyright works which become part of public domain such as extracts of judgments or statements of case, or client architectural drawings available through planning portals, may be affected by any proposed exceptions permitting their use to train AI models. 

Licensing of copyright material and its interaction with AI model training

Copyright holders currently monetise their protected works through licensing arrangements. Licensing of works for use with AI, whether for training or other uses, can also give creators opportunities to generate new sources of income. The market for licensing copyright works for use with AI technology is still new and evolving, but practices are market-led with limited regulation. 

Technical standards are also developing, including those on the use of web crawlers and AI agents. The UK law, unlike USA, Japan, China and some other jurisdictions, has only very limited exceptions to copyright for text and data mining (TDM) in non-commercial circumstances. TDM processes are often significant in facilitating increased AI use and development. 

The licensing market continues to grow to the benefit of right holders and AI developers, though limited public information on licensing deals makes it difficult to fully assess their impact. Some rights holders license their works to allow TDM, but others do not. This has financial costs for people using data mining software, and has led to criticism that investors would choose to support AI and TDM projects in other countries because of the current limited UK framework. The government is therefore giving more thought to this interaction, but has (for the moment) ruled out the previously-proposed exception to TDM processes.

Legal protection given to CGWs

The UK is internationally recognised for its strong framework for enforcing intellectual property (IP) rights, but AI may pose new challenges for enforcement. 

The UK is one of only a handful of countries that has statutory provision for the authorship of computer-generated works (CGWs). Section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA) provides: “In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.”

Section 178 of the CDPA defines “computer-generated”, in relation to a work, as meaning that the work is generated by computer in circumstances such that there is no human author of the work.

The statutory definition of CGWs is not clear. There is clearly scope for interpretation of authorship depending on the degree of input by computer and the skill, labour and intellectual creation of the of the person inputting the data. In the context of a word-processed document, it is clear who the author of a document is. The creator of the software does not supply the necessary ingredient of originality required for composition of the document. Instead, the user does. If AI assists a user to take a photograph, provided the photograph expresses the creativity of the photographer, it would be protected as an artistic work, regardless of whether AI assisted them. However, AI has now developed to the stage where CGWs can be produced in large quantities without the same challenge or cost, in which case the courts have tended to find an author as either the user or the developer. 

It has also been argued that the statutory definition of CGW provided by s9(3) is inherently contradictory with regard to originality in CGWs, since the legal concept of originality is defined by reference to human authors and characteristics such as personality and skill. Under Section 1 of the CDPA, in order for copyright to subsist in a literary, dramatic, musical and artistic (LDMA) work, it must be “original”. THJ Systems Ltd v Sheridan [2023] EWCA Civ 1354 (as applied in Lidl Great Britain Ltd v Tesco Stores Ltd [2024] EWCA Civ 262) confirms the latest test for originality is “author's own intellectual creation.”

 This test requires that the author was able to express their creative abilities in the production of the work by making free and creative choices so as to stamp the work created with their personal touch. This contradiction has led some to question whether the provision could ever apply in practice.

If one focuses on the "author's own intellectual creation" test , it is difficult to see how originality could be located if a work is deemed to have no human author in accordance with Section 178 so that it falls into the CDPA's CGW category. Even where a work is deemed to be a CGW, the courts have shown a willingness to find a human author (either the user or AI system developer) who has supplied the necessary originality. There is always however in principle a class of works in respect of which this is not possible. 

The rationale for express CGW protection was sensible in a time where the technology relating to computer-generated works was in its (relative) infancy and content such as statistical data, weather maps, and outputs from expert systems was considered to be valuable and worthy of protection. However, its limited application in practice has led to questions as to whether it achieves Parliament’s intention. 

The UK government’s AI and copyright predicament 

As utilisation of AI continues to gain traction, the need for clarity on AI usage in training models and CGW authorship is ever-increasing. The government have launched a series of consultations including those in November 2021 and in December 2024. The government’s aims are “unlocking the extraordinary potential of AI to grow the economy and improve lives”, and it made proposals in December 2024 which would potentially allow tech companies to use copyrighted works to train their models under a TDM process, with an opt-out option for original authors of the copyrighted works (the “December 2024 Proposals”). 

Various industries presented immediate backlash, including particularly the music industry with Sir Paul McCartney, Dua Lipa, Sir Elton John and others voicing their criticisms of the December 2024 proposals. The government have recently (March 2026) rowed back from the December 2024 Proposals, stating that the government "no longer has a preferred option" for what to do next, and will take more time to “get it right.” 

TechUK’s response urges caution stating “the UK risks drifting further behind other major AI economies. Countries including the United States, Japan, Singapore and the European Union have already moved faster to provide legal clarity around text and data mining, creating more predictable environments for AI development and investment.” Meanwhile, UK Music Chief Executive Tom Kiehl said the music industry should be able to work “without the constant fear that the fruits of their labour could effectively be taken by AI firms without payment or permission.”

As the government wrestles its ambition to unlock growth and innovation whilst protecting human creativity, clearly similar issues may become prevalent in other industries. For example, the protection of copyright in construction designs and drawings has long been a major concern for architects. Similarly, copyright issues can arise in the work of legal practitioners. Both professions have to address the scope for their ‘public’ documents to be accessed by TDM and then used to train models, potentially at the expense of the copyright holder and in breach of any protective notices prohibiting copying and/or licensing agreements. Similarly, the definition of CGW is only likely to come under increased scrutiny as technology develops which may affect advice given to clients producing works with AI assistance.

Uncertainty and change

It doesn’t appear that reform is as imminent as it seemed when the December 2024 Proposals were published. However, the pace of change presented by AI training is unrelenting and pervasive. 

Lord Holmes of Richmond MBE (one of the UK’s leading authorities on digital regulation) is quoted as follows: “Whether creative or citizen, developer or deployer, clarity is required. The UK needs a cross economy cross-sector AI Bill and the time for this is well overdue.”

Some form of reconciliation between AI utilisation (with disclosure of type of data used to train models) and creative protection (through licensing) appears inevitable. Practitioners will need to look at any reform that may be implemented carefully, so that copyright material can be adequately protected and appropriate advice can be given to client tech companies using AI and/or creatives seeking to protect works.