UK AI Copyright Laws: CDPA s.9(3) and What It Covers

The United Kingdom holds a rare distinction in global copyright law: under section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA), the UK can grant copyright protection to works generated entirely by a computer, with no human author required. Most major economies offer no equivalent route.

Information last verified on 2026-09-30. This article presents general legal information, not legal advice. UK policy on AI and copyright is under active review.

Scope: This article covers United Kingdom copyright law as it applies to AI-generated output, AI training and text-and-data mining, and software protection under the CDPA 1988. For a global comparison, see how AI and copyright differ worldwide.

Computer-Generated Works: The Section 9(3) Route

The UK is unusual among major economies in providing a statutory path to copyright protection for works generated by a computer without a human author. Under section 9(3) of the CDPA 1988, the author of a computer-generated literary, dramatic, musical, or artistic work is "the person by whom the arrangements necessary for the creation of the work are undertaken." Section 178 of the CDPA defines "computer-generated" as a work generated by computer in circumstances where there is no human author.

This means that if a developer, business, or individual sets up and runs the system that produces the work, they may qualify as the statutory author. The protection exists independently of any creative contribution by a human to the final output, at least on the face of the statute.

The term of protection is shorter than for human-authored works. Under section 12(7) CDPA, a computer-generated work is protected for 50 years from the end of the calendar year in which it was made. Standard works authored by a natural person attract the EU-legacy life-plus-70-year term.

This route may not last. In its March 2026 report to Parliament on copyright and AI, the government said that, in the absence of evidence of its ongoing value, it proposes removing the specific protection for wholly computer-generated works, while copyright should continue to protect works created with AI assistance. That change would need legislation; until it is made, section 9(3) and the 50-year term remain in force.

Tension with the Originality Standard

Section 9(3) was drafted in the 1980s with automated database generation in mind, long before modern generative AI. Applying it to large language models and image generators raises a question the statute does not resolve: does prompting an AI system constitute "making the arrangements necessary" for the work?

No UK court has ruled directly on whether a prompt is sufficient arrangement. The question is further complicated by the originality standard the UK now applies. Following Brexit, the Court of Appeal confirmed in THJ Systems Ltd v Sheridan [2023] EWCA Civ 1354 that the UK requires an "author's own intellectual creation" for a work to qualify as original, a test imported from EU case law before the UK's departure. This standard is difficult to satisfy where no human exercises intellectual judgment over the form of the output.

The result is a structural tension: s.9(3) provides a mechanism for authorship without a human, but the originality requirement may demand a human intellectual contribution the prompt alone cannot supply. No UK court has yet resolved this tension. The leading UK generative-AI copyright judgment so far, Getty Images v Stability AI (discussed below), concerned model training and the model itself, not who owns AI output.

Text-and-Data Mining for AI Training

Training a generative AI model typically requires copying large volumes of third-party text, images, or audio to extract patterns. Under UK copyright law, that copying is potentially infringing unless a licence or exception applies.

Section 29A of the CDPA permits text-and-data mining (TDM), but only within strict limits:

Commercial AI developers training on publicly available web content cannot rely on s.29A. They require either a licence from rights holders or must argue that their activity falls within some other exception, none of which straightforwardly covers commercial model training at scale.

The first UK judgment on generative AI and copyright did not settle whether training in the UK infringes. In Getty Images v Stability AI [2025] EWHC 2863 (Ch), decided on 4 November 2025, Getty abandoned its claim over training because it accepted there was no evidence that Stable Diffusion was trained or developed in the UK. The High Court held that a model can in principle be an 'article', including an intangible electronic copy, but dismissed Getty's secondary infringement claim because Stable Diffusion was not an infringing copy of its works.

In December 2025 the court gave Getty permission to appeal on that point ([2025] EWHC 3343 (Ch)), and the government's March 2026 report describes the judgment as subject to appeal.

The Reform Process: 2022 to 2026

The current restriction on commercial TDM has been the subject of sustained policy debate. A proposed broad TDM exception announced in 2022 was shelved after industry opposition from creative sectors. A government consultation running from December 2024 to February 2025 floated an opt-out model, under which rights holders could reserve their works from AI training through a machine-readable signal, while permitting training by default where no opt-out was lodged.

Sections 135 to 137 of the Data (Use and Access) Act 2025 required the Secretary of State, not AI developers, to publish an economic impact assessment of the consultation options and a report on the use of copyright works in developing AI systems, and to lay both before Parliament. The Act placed no transparency or reporting duty on AI developers, and it did not amend copyright law or create a commercial TDM exception.

The government published the report and impact assessment on 18 March 2026. Citing strong views in the consultation and gaps in the evidence, the report said a broad copyright exception with opt-out is no longer the government's preferred way forward, and proposed gathering further evidence on how copyright law affects AI development. It also proposed removing the specific protection for wholly computer-generated works, while keeping copyright for works created with AI assistance.

No legislation has yet been enacted to change the TDM position. Commercial model training without a rights-holder licence remains legally uncertain at best.

Software Protection

AI systems are themselves software, and copyright protection for the underlying code is straightforward under UK law. Section 3(1) of the CDPA classifies computer programs as literary works, giving them the same protection as any text. This covers source code, object code, and preparatory design material.

Protecting an AI model as software is therefore well established. The harder questions concern the outputs the model produces, addressed above, and the training data that shaped its weights, addressed in the TDM section.

How the UK Differs from the United States

The UK and the US represent contrasting positions on AI-generated copyright:

The UK's s.9(3) route is the most significant difference. A business that builds and operates a generative AI system in the UK has at least a statutory basis to claim authorship of outputs. In the US, no equivalent exists, and copyright in AI-generated content is denied on the face of Copyright Office policy.

This article presents general legal information about UK copyright law and AI. It is not legal advice. The law in this area is developing quickly. If you need advice on your specific situation, consult a UK-qualified solicitor. Information current as of 2026-09-30.

Related articles

• How AI and copyright differ worldwide
• AI copyright law in the United States
• European Union AI copyright laws
• Australia AI copyright laws

Last updated: 2026-09-30.

Frequently Asked Questions

Can AI-generated content be protected by copyright in the UK?

Possibly, through the specific route in section 9(3) of the CDPA 1988. The person who makes the arrangements necessary for the creation of a computer-generated work is treated as the statutory author. However, the 'author's own intellectual creation' originality standard may also apply, and no UK court has yet ruled on whether a prompt satisfies that standard for modern generative AI output. The position is legally uncertain.

How long does copyright last for AI-generated works in the UK?

Under section 12(7) of the CDPA 1988, a computer-generated work is protected for 50 years from the end of the calendar year in which it was made. This is shorter than the standard term for human-authored works, which lasts for the author's life plus 70 years. In March 2026 the government proposed removing the specific protection for wholly computer-generated works, but that would need legislation, and s.9(3) and the 50-year term remain in force until then.

Can AI companies train their models on copyrighted material in the UK?

Not without a licence if the purpose is commercial. Section 29A of the CDPA permits text-and-data mining only for non-commercial research by a person with lawful access to the material. No commercial TDM exception is in force, despite several years of government consultations on the topic.

What did the Data (Use and Access) Act 2025 change about AI copyright?

It did not change the copyright rules. Sections 135 to 137 of the Data (Use and Access) Act 2025 required the Secretary of State to publish an economic impact assessment and a report on the use of copyright works in developing AI systems, which the government did on 18 March 2026. The Act placed no transparency or reporting duties on AI developers, did not create a commercial TDM exception, and did not otherwise amend the copyright provisions of the CDPA 1988.

Is the UK planning to introduce a commercial text-and-data mining exception?

The government has been consulting on the question since 2022 without reaching a legislative outcome. A broad TDM exception was proposed in 2022 then shelved. A December 2024 to February 2025 consultation proposed an opt-out model. A March 2026 government report said a broad exception with opt-out is no longer the government's preferred way forward and proposed gathering further evidence rather than legislating. No commercial exception exists.

How does the UK approach to AI copyright compare with the EU?

The EU's Directive on Copyright in the Digital Single Market introduced a commercial TDM exception for AI training, subject to an opt-out for rights holders. The UK, post-Brexit, did not implement that directive. The UK retains the non-commercial-only restriction under s.29A CDPA and has not enacted a comparable commercial exception, leaving UK and EU law on this point diverged.