AI and Copyright: Who Owns AI-Generated Content in the UK?

Who owns a piece of writing generated by ChatGPT? Can a business claim copyright over an AI-created logo? What happens when an AI’s training data included your book, your photographs, or your music? These questions are landing in UK boardrooms and creative studios right now — and the answers are messier than most people expect.

UK copyright law was written decades before generative AI existed. The Copyright, Designs and Patents Act 1988 contains a provision about “computer-generated works” that was ahead of its time in 1988 but was designed for basic software, not large language models trained on half the internet. Today, with AI generating images, articles, code and audio at industrial scale, the legal framework is being stretched to its limits.

This guide explains where UK law currently stands on AI and copyright, what the Intellectual Property Office is doing about it, and what businesses and creators need to know right now.

What Counts as AI-Generated Content?

AI-generated content is any output produced primarily by an artificial intelligence system, where the human’s contribution was mainly operating the tool rather than expressing creative ideas. This includes images from Midjourney and DALL-E, written text from GPT-4 and Claude, music from Suno and Udio, and video from Sora and Runway.

The legal questions get complicated fast because there is a spectrum. A novelist who uses AI to generate rough chapter outlines and then writes the actual prose has a strong creative claim. A marketing team that types a two-word prompt and publishes the output unedited has almost no creative input at all.

UK courts have not yet ruled definitively on where on that spectrum copyright protection kicks in. This ambiguity is creating real problems for businesses building AI-driven content operations. In 2026, an estimated 60% of UK marketing teams use AI for content creation in some form.

The UK’s Unique Position on Computer-Generated Works

Most countries have no explicit provision for AI-generated content. The United States, the European Union and Australia all treat copyright as requiring human authorship. Works produced entirely by machines cannot be registered or protected in those jurisdictions.

The UK took a different path in 1988. Section 9(3) of the Copyright, Designs and Patents Act says that for a computer-generated work, the “author” is “the person by whom the arrangements necessary for the creation of the work are undertaken.” This is a unique provision in global copyright law.

In theory, this means the person who operates an AI tool and commissions its output can be treated as the author. In practice, it raises enormous questions: Does typing a prompt count as “making the arrangements”? Does the AI company that built and trained the model have a claim? Does the section even apply to probabilistic AI outputs, which were unimaginable in 1988?

These questions have not been tested in a UK court in the context of modern generative AI. Until they are, anyone relying on Section 9(3) to protect AI-generated commercial content is building on uncertain ground.

The Originality Problem

Even if a human is deemed to be the “author” of AI-generated content under Section 9(3), there is a second hurdle: originality. UK copyright law requires a work to reflect the “own intellectual creation” of the author — a standard derived from EU law and confirmed in UK case law.

For a work to meet this test, it must involve some creative choices. A photograph taken by a human involves choices about framing, lighting and timing. A poem written by a human involves choices about words, metre and imagery. Does a prompt typed into an AI meet this test?

Longer, more detailed prompts that specify particular visual or stylistic choices are more likely to pass the originality test. Short, generic prompts — “write a blog post about coffee” — almost certainly do not. Businesses that want to protect AI-generated content should invest time in prompt crafting and retain records of their creative process.

The Training Data Problem: Who Infringed What?

A separate legal battle is being fought over the data used to train AI systems. Large language models and image generators are trained on billions of items of text, photographs, music and other content. The vast majority of this content is copyright protected.

The AI industry has argued this falls within existing exceptions for text and data mining (TDM). The creative industries have argued it does not, and that AI companies should pay licences for training data just as broadcasters pay licences for music.

In the UK, Getty Images filed a High Court claim against Stability AI in 2023, arguing that millions of Getty photographs were scraped without permission to train Stable Diffusion. The case is making its way through the courts; a full trial is expected no earlier than 2027.

The UK Intellectual Property Office consulted on whether to introduce a broader TDM exception that would allow AI companies to train on any publicly available content. Following fierce opposition from the Society of Authors, the Publishers Association and the BBC, the government stepped back from introducing this exception. The debate is far from over.

What the Intellectual Property Office Is Doing

The Intellectual Property Office has been the most active government body on AI and copyright. Since 2021 it has run multiple consultation exercises and published detailed reports. Its position as of 2026 is roughly this: Section 9(3) is being maintained for now, but the law may need updating as AI technology evolves and courts start ruling on specific cases.

The IPO has also published non-binding guidance for businesses. It suggests that businesses document their use of AI tools — which tools they used, what prompts they entered, what editing they carried out — so they can demonstrate human creative input behind any content they want to protect.

The IPO guidance is useful but it is not law. It reflects where the IPO thinks the law is, not where a court has confirmed it to be. Treat it as a sensible starting point, not as legal certainty. Any business with significant IP exposure should seek specialist counsel.

Practical Implications for UK Businesses

For UK businesses using AI to generate content, there are several practical steps that reduce legal risk. First, document everything. Keep records of prompts, tools used, dates and any editing or curation of AI output. This builds evidence of human creative contribution that strengthens any future copyright claim.

Second, check terms of service. Most major AI tools assign generated content to the user in their terms of service. OpenAI, Anthropic and Adobe all do this. Read these carefully and understand what rights you are actually getting and what restrictions apply to commercial use.

Third, prefer AI tools trained on licensed data where copyright is commercially important. Adobe Firefly, for example, is trained exclusively on Adobe Stock and openly licensed content, significantly reducing training-data infringement risk. This matters for businesses in regulated industries where IP integrity is audited.

Fourth, do not rely on AI-generated content for your core IP assets without legal review. Trademarks, patents and design rights each have different requirements. A brand identity built entirely on AI-generated logos may have weaker legal protection than one designed by humans.

Fifth, seek specialist legal advice before commercialising AI-generated content at scale. IP law firms have built practices around this question. The cost of a brief consultation is small compared to the cost of a copyright dispute.

What UK Creators Need to Know

For individual creators — writers, photographers, illustrators and musicians — the AI copyright debate has direct consequences for income. If AI systems trained on creative works can produce competing outputs without paying the original creators, the economics of many creative professions become harder to sustain.

The good news is that the UK government has heard these concerns. The TDM exception that would have allowed unlimited scraping was not introduced. Licensing frameworks — where AI companies pay royalties for training data, similar to the music licensing model — are gaining political support across both major parties.

The bad news is that enforcement is slow, the law is unclear and AI companies are moving far faster than regulators. Creators who rely on their work commercially should register their key works with copyright registries where possible, document their creative processes and follow the IPO’s guidance as it develops.

Creators who use AI as a tool in their own work are in a stronger legal position. A photographer who uses AI to speed up editing retains copyright over the finished work, because the creative choices — what to photograph, when, how to frame it — remain theirs. A musician who uses AI to generate a backing track and then performs and mixes it themselves has a defensible authorship claim over the final recording.

What This Means for UK Investors and Businesses

The AI and copyright question has significant financial implications. Businesses that generate content using AI and treat it as fully protected intellectual property are taking a legal risk. That risk may not materialise — but it could affect valuations in due diligence, licensing deals and future litigation exposure.

For investors in UK AI companies: the unresolved training data question remains a genuine liability. The Getty vs Stability AI case, and others like it, could result in significant damages awards or forced licensing arrangements that fundamentally affect the economics of AI content generation.

Watch the IPO consultations closely over the next 12 to 24 months. A legislative response is coming — the only question is what shape it takes. When I look at the political trajectory, a licensing model rather than a blanket exception seems more likely. Businesses that build for a licensed future will be better positioned than those who assume free access to training data forever.

This article is for educational purposes only and does not constitute legal or financial advice.