Copyright and AI in the UK: The Battle is Far from Over

An emblematic lawsuit that delivers little

The case was expected to set a precedent. Getty Images largely lost its lawsuit against Stability AI in the High Court of London, with Judge Joanna Smith only recognizing partial trademark infringement while dismissing secondary copyright infringement ^[2]. A stinging setback for the photo agency, which accused the publisher of Stable Diffusion of scraping millions of its images without permission to train its image generator.

The outcome was predictable long before the verdict. In the midst of final arguments, Getty had dropped its main infringement claims, significantly reducing the scope of one of the world's most scrutinized disputes over the training of generative AIs ^[3]. A partner at the EIP firm explained that this tactical retreat was likely due to Getty's inability to demonstrate a sufficient link between the model's training and the UK territory ^[6] — most of the operations took place in the United States.

The crux of the matter: where was the training conducted?

This territorial question is not a technical detail. It reveals a gaping flaw in the UK's legal arsenal: how to apply national law to models trained abroad but marketed in the UK? Intellectual property law specialists agree that the current legal framework, designed long before the explosion of generative AI, is inadequate to protect creators ^[1]. The debate flared up with the launch of ChatGPT in November 2022, with artists, photographers, and writers collectively denouncing the unauthorized use of their works to feed algorithms that could then directly compete with them.

The government backs down in the face of creators' revolt

Politically, the UK government initially attempted to impose an exemption allowing AI companies to train their models on protected works without prior authorization. The backlash was spectacular: key parliamentarians joined the protest, and all major British newspapers published identical front pages denouncing the project ^[4]. Figures like Kate Bush and Annie Lennox even went as far as releasing an album of silence, symbolizing what creation would become without protection.

Personalities as diverse as Stephen Fry, Paul McCartney, and Dua Lipa also penned an open letter to the Times, insisting that the British creative industries are willing to participate in the AI revolution, but not without a solid intellectual property foundation ^[12]. As a result, faced with this unprecedented mobilization, the government eventually abandoned its initial reform plan, a notable retreat that does not, however, close the debate ^[11].

A consultation that resolves nothing

To try to break the deadlock, the government launched a wide-ranging public consultation, which closed in February, gathering more than 11,500 responses ^[8]. Notably, the most favored option was not the relaxation sought by tech giants, but rather a stricter framework, requiring a license for any AI development, including models trained outside the UK ^[9].

The interim report published in December confirms what everyone suspected: positions remain completely polarized, pitting creative industries against AI developers, with no satisfactory compromise in sight ^[10]. A new government report, accompanied by an impact study, is now expected by March 18, 2026 — yet another deadline in a case that has dragged on for over three years.

When AI-generated art goes to auction

Beyond the courts and Parliament, the controversy has also played out on the symbolic field of art. Christie's auction of AI-generated works sparked a wave of protests, with more than 6,000 artists signing a petition to cancel it ^[7]. The auction house held its ground despite the pressure, illustrating how the art market is now also caught up in this battle over the very definition of creativity.

This philosophical question is not legally trivial: British law has long recognized, via the famous section 9(3) of its copyright law, the possibility of attributing a computer-generated work to the person who enabled its creation — although this provision, several decades old, was not designed for modern generative AIs capable of producing photorealistic images in seconds.

What could change in 2026

Several options are on the table to get out of the rut:

• A mandatory licensing regime for any AI model marketed in the UK, regardless of the training location;
• Greater transparency required from companies on the data used for training;
• A collective remuneration mechanism for creators, inspired by existing systems in music;
• Maintaining the current status quo, deemed insufficient by almost all consulted stakeholders.

Pending legislative clarification, the judicial route continues to be tested on a case-by-case basis. Other disputes could follow Getty's example, with the hope this time of better establishing the necessary territorial link to succeed in a copyright infringement claim in UK courts.

A debate far from closed

Three years after ChatGPT burst into public debate, the UK has yet to resolve the fundamental question: who owns the rights to a creation from an algorithm fed with human works? Between a government caught in a bind, tech giants eager to move forward, and creators determined not to yield, the March 18, 2026 deadline promises to be another round, likely not the last, in this tug-of-war with significant economic and cultural stakes.