Legal Guide Pro

AI and copyright in the UK: what the law says now

News & Trends4 min read5 May 2026

By · Published

Where UK law stands on AI and copyright in 2026: training AI on copyrighted works, the text and data mining exception, Getty v Stability AI, who owns AI outputs, the government's March 2026 report, and practical tips.

Generative AI raises difficult copyright questions: can AI developers train models on copyrighted material, who owns content an AI creates, and can AI outputs infringe? UK law hasn't yet been changed specifically for AI, and the position is still developing. This guide summarises the current state of play in 2026. It isn't legal advice.

Training AI on copyrighted works

Training an AI model usually involves copying large amounts of material. Under UK law, copying a substantial part of a protected work without permission can infringe copyright unless an exception applies.

The UK has a text and data mining (TDM) exception, but it only covers computational analysis for non-commercial research, by people with lawful access to the work. Commercial AI training isn't covered.

The government's position

In December 2024, the government consulted on introducing a broader TDM exception allowing commercial AI training unless rights holders opted out. The proposal drew strong opposition from creative industries. The Data (Use and Access) Act 2025 required the government to report on copyright and AI, and it published its report and impact assessment on 18 March 2026. The government dropped its preferred opt-out exception and didn't adopt a replacement; it said it would gather more evidence and consider alternatives. For now, the existing law applies.

Getty Images v Stability AI

In November 2025, the High Court ruled in a closely watched case brought by Getty Images against Stability AI, developer of the Stable Diffusion image generator. Getty dropped its main training claims during the trial, partly because the training took place outside the UK. The court rejected Getty's secondary infringement claim, holding that the trained model's weights were not an "infringing copy" because they didn't store or reproduce Getty's images. Getty succeeded only on limited trade mark points about watermarks in some outputs. The case leaves many questions open, particularly about training carried out in the UK.

Who owns AI-generated content?

UK law has an unusual provision for "computer-generated works" with no human author: copyright belongs to the person who made the arrangements necessary for the work's creation, and lasts 50 years. How this applies to modern generative AI is uncertain, and the government has consulted on whether to keep it. Where a person uses AI as a tool but makes creative choices, the result may be protected as their own work in the usual way.

Can AI outputs infringe?

Yes, potentially. If an AI output reproduces a substantial part of a protected work, such as a recognisable character, lyrics or a photograph, using or publishing it could infringe, regardless of how it was created. Outputs can also raise trade mark or passing-off issues.

Practical tips for creators and businesses

Other countries

The EU's AI Act requires general-purpose AI providers to respect copyright opt-outs and publish summaries of training data. In the US, numerous court cases are testing whether training is "fair use". These developments affect UK businesses using international AI services.

Frequently asked questions

Can I copyright an image I made with AI?

Possibly, under the computer-generated works provision or if you made significant creative contributions, but it's uncertain.

Can I use AI to write content for my business?

Yes, but check the tool's terms, review outputs, and take responsibility for what you publish.

Sources and further reading

Facts checked on 20 October 2026. Spotted an error? Tell us and we will correct it.

Get the next guide first

New articles land every week. Follow along by RSS or bookmark the homepage.

Follow by RSS