What The UK’s Data Bill Means For Music, AI And Copyright

After months of political turbulence, the UK’s Data (Use and Access) Bill has finally passed Parliament. Marketed as a major update to the country’s data infrastructure and digital governance, the bill covers everything from NHS data interoperability to digital ID systems and AI-enabled decision-making. The text is broad in scope, modernizing the UK’s GDPR, streamlining data subject access, and enabling more fluid data sharing across public services and smart infrastructure.

However, it also weakens restrictions on automated decision-making and sidesteps key copyright issues raised by AI.
Although the Data Bill does not legislate on copyright directly, creative industries had hoped it would include minimal safeguards for the use of copyrighted works in AI training.

In parallel to the bill, the government has signaled, through its consultation on generative AI conducted by the UK Intellectual Property Office, support for a model that would allow AI developers to mine copyrighted content by default, unless rights holders explicitly opt out. This mirrors the EU’s controversial text and data mining (TDM) exception, a proposal that many in the creative industries see as deeply problematic.

Attempts to introduce a transparency duty for AI developers were proposed and passed repeatedly in the Lords, but were ultimately rejected in the Commons for the sixth and final time.

For the UK’s creative sectors, music, publishing, film, and visual arts, which collectively generate over £124 billion annually, the final version of the bill represents a missed opportunity and a potentially dangerous precedent. It leaves songwriters, recording artists, and rights holders unable to determine whether their work has been ingested into AI training datasets, with no clear obligation for companies to provide transparency or seek permission.

Training Without Traceability

The Lords amendment, proposed by Baroness Beeban Kidron, became the focal point of this battle. Her proposal was straightforward: require AI developers to disclose what datasets and copyrighted material they used to train generative AI systems. The amendment kept passing in the Lords with growing support, only to be killed repeatedly in the Commons. The government refused to accept it, claiming that it would stifle innovation and that copyright would be addressed in a separate AI-specific bill after a public consultation.

But as Jane Clementson, a lawyer who advises media and creative businesses on the creation and exploitation of intellectual property, explains, the government had a ready excuse: “The DUA Bill was never intended to address copyright law, so amendments about AI training data were resisted. The Government’s view was that this wasn’t a copyright bill—wrong vehicle for such a complex issue.” This reasoning allowed ministers to sidestep the core issue while promising to address it later in a separate AI bill.

The current UK copyright framework under Section 29A of the Copyright, Designs and Patents Act allows TDM only for non-commercial research purposes. This means AI developers may lawfully copy and analyze copyrighted content only if the use is non-commercial, and even then, only under specific conditions.

The Data Bill does not change this legal provision. However, it fails to strengthen copyright protections or clarify enforcement, despite the rapid growth of commercial AI training models.

“Support for innovation shouldn’t come at the cost of fairness,” explains Rick Gleave, a music-tech strategist and founder of Music Foundry. “The current trajectory risks building AI systems on the backs of unlicensed creative works, music, lyrics, performances, without attribution or compensation. That’s not a sustainable model.”

Because the law lacks meaningful enforcement and does not mandate dataset disclosure, AI developers can ingest massive libraries of music and argue that their use remains “non-commercial” as long as they don’t sell the original content directly. Instead, they train generative models that create synthetic outputs which compete directly with the original works, often replicating stylistic, lyrical, or sonic elements. And the value extracted at the training stage powers downstream applications and services generating vast profits.

“The refusal to amend the bill weakens the UK’s standing as a defender of copyright and the creative industries,” argues Gleave. “We’ve traditionally prided ourselves on striking a fair balance between innovation and rights protection, but this bill tips the scales toward data access and AI development without adequate safeguards for creators.”

The asymmetry is stark. Developers gain free rein to mine cultural data while creators remain in the dark. The proposed opt-out mechanism might sound like a compromise, but without mandatory transparency, it becomes meaningless in practice. Rights holders cannot opt out of training datasets they aren’t even aware they’re part of.

Clear Law, Rampant Violations

These concerns aren’t theoretical. The International Confederation of Music Publishers (ICMP), the global trade body representing Majors, Indies and 80 different national trade associations across 6 continents has documented clear evidence that commercial AI systems, including Suno, Udio, Gemini, and DeepSeek, have been trained on unlicensed music. When prompted, despite claims of safeguards, these systems can generate synthetic outputs that replicate the sonic and lyrical fingerprints of songs despite no licensing agreements being in place.

The legal requirement is clear: AI developers must license copyrighted material when using it for commercial purposes. The problem arises because enforcement has fallen behind and multiple lawsuits show just how blurred the lines have become. Yet the following examples collected by ICMP contradict government claims that enforcement is premature, showing that unlicensed reproduction is already widespread and increasingly sophisticated.

– When prompted to analyze the lyrics of “Billie Jean” by Michael Jackson, Google’s Gemini model outputs the full lyrics, despite no licensing agreement with rights holders. This directly contradicts claims made by some AI developers and policymakers that generative systems are trained only on “non-consumptive” data or that robust filters are in place to prevent reproduction of copyrighted content.

– DeepSeek, a Chinese-developed model, goes even further. It can reproduce full copyrighted lyrics, including recent songs, formatted and tagged with metadata scraped from platforms like Spotify. This suggests an intentional bypass of standard licensing practices and highlights how easily some developers evade rights protections.

– In Germany, the collection society GEMA has flagged a Suno-generated track for strong similarities to Alphaville’s 1984 hit “Big in Japan.” According to GEMA, the AI-generated version reproduces the lyrics almost verbatim, with matching phrasing and structure, despite no licensing deal existing between Suno and Alphaville’s rights holders. GEMA has filed a lawsuit against Suno for copyright infringement, further alleging that the company has reproduced protected works without permission across jurisdictions. The suit also cites additional tracks allegedly copied from Alphaville (Forever Young), Kristina Bach (Atemlos durch die Nacht), Lou Bega (Mambo No. 5), Frank Farian (Daddy Cool), and Modern Talking (Cheri Cheri Lady).

gema.deAudio samples Suno – gema.de

– Similarly, Udio, a fast-growing AI music generator, has been shown to produce songs that closely imitate the Beatles’ musical style, lyrical tone, and even vocal timbre. Prompts like “write a Beatles-style ballad about longing” yield tracks that mirror the harmonic structure, instrumentation, and production techniques of Lennon–McCartney compositions. While not replicating lyrics verbatim, the outputs often share thematic content, rhyming patterns, and arrangements, effectively creating derivative works. Udio has no licensing agreement with Apple Corps, Sony/ATV, or any entity managing the Beatles’ catalog, making these outputs clear examples of unlicensed stylistic appropriation. (This example is drawn from the ICMP evidence submission).

“It actually doesn’t need to work this way,” says John Phelan, CEO of ICMP. “Ours is an industry built on exclusive rights, and what that literally means is not so much that we want to restrict use, some other creative sectors are much less willing to license works, but rather that commercial users need prior authorization to be legal.”

The irony runs deeper. ICMP’s analysis of tech company contracts reveals a telling double standard. Google, Microsoft, Meta, OpenAI, Suno, Udio and others all include clauses demanding “no use of our content without express prior written permission.”
Calling out the contradiction, Phelan says: “We in the music industry should not be reluctant to point out this commercial hypocrisy and demand total respect for our songwriters’ property rights.”

Effectively, the UK’s copyright landscape has no lack of clarity; rights holders already have control and licensing training data for commercial gen AI is required by law. But Science and technology secretary Peter Kyle mentioned several times that his preferred outcome on AI and copyright is a new copyright exception allowing unlicensed training and that UK copyright law is currently uncertain, suggesting protesting artists are just ‘people who resist change’.

The Politics of AI-First Growth

To understand how Britain reached this point, look beyond the legislation to the political backdrop. Labour won a commanding 403 out of 650 MPs but with only 33% of the national vote, strong legislative control built on fragile public support. Desperate for economic wins, the Starmer government has bet heavily on positioning the UK as a global AI hub. The Data Bill serves that agenda: deregulate, incentivize, and let innovation flourish. “By rejecting the Lords’ transparency amendments and deferring copyright enforcement to a vague future bill, the government has effectively given AI developers free fuel in the form of unlicensed cultural content” observes Jake Beaumont Nesbitt, Consulting Artist Manager and advisor on entertainment tech, director of Innovation at International Music Managers Forum. But he also notes that the UK Government just backed the music industry with a £30 million investment package and adds “One could see this as a bunch of flowers offered to the creative industries after the Government ran off with the Tech Sector.”

Ministers frame the bill as a growth engine, freeing NHS staff from admin and energizing fintech services. Copyright, they argue, is too complex to address in legislation aimed at “improving people’s lives.”

But peers openly accused the government of bowing to Big Tech lobbying. New figures obtained by Democracy for Sale reveal that Labour ministers and senior civil servants met with tech industry executives and lobbyists an average of six times a week during the government’s first six months in office. Peter Kyle asked Google’s head of AI, Demis Hassabis to “sense check” AI policy and Hassabis is now a formal advisor on the government’s AI plan. The Technology Secretary also said he would “advocate” for Amazon at the UK’s competition regulator and their case against Amazon was then dropped.

As Baroness Kidron put it: “Silicon Valley has persuaded the government that it’s easier to redefine theft than make them pay for what they have stolen.” And Phelan confirmed “Artificial Intelligence covers a multitude of different services. It’s as specific as using the word technology. But suffice to say the music industry has been longstanding adopters of and innovators in AI, from licensing admin to searching song databases, copyright infringement prevention to amazing new visual effects to enhance the concert experience. Any characterisation of AI and the music industry as antiphonal is way wide of the mark.”

The Promise of Tomorrow

Technology Secretary Peter Kyle has promised a “comprehensive” AI bill that will revisit transparency and opt-out mechanisms, potentially arriving by May 2026. The government committed to publishing reports on copyright and AI within nine months, including analysis of economic impacts on creators and developers.

Jane Clementson explains: “The Secretary of State is obliged, within nine months of Royal Assent being given, to publish an impact assessment of the economic impact of each of the four policy options described in the government’s recent Copyright & AI consultation paper — including the impact on copyright owners and development users.” “They must also present a report to Parliament on how copyrighted works are being used to develop AI systems,” she adds.

But creative industries worry this delay is strategic. And as Beaumont Nesbitt points out: “Rather than a well-thought-out long-term strategy, this hands-off approach is a short-term political gamble. The Government believes it’s too soon to regulate, and is giving these (mostly ex-UK) companies not only a green light, but free fuel.”

Each month of delay invites further ingestion of Britain’s cultural catalogue, expanding AI libraries at zero cost while eroding the scarcity on which copyright economics rests. As Nick Breen, Partner at the global law firm Reed Smith LLP, explains, here’s what to expect next: “Now that the government has committed to providing a report, we can expect intense lobbying from both sides—on everything from transparency and copyright exceptions for training, to international interoperability, protection of AI outputs, and image rights. Given the UK’s hesitation to legislate prematurely, it now faces pressure to offer clarity and show how it has balanced competing interests. In the meantime, ongoing litigation—such as Getty’s case against Stability AI in the High Court—will continue to shape the landscape.”

But by the time comprehensive legislation arrives, the market may have already normalized unlicensed training. John Phelan makes clear that: “To date, I have still not seen any provable pathway to becoming more economically competitive by way of a government reducing copyright standards. There is no credible evidence that if you make that industrial policy choice, an increased influx of foreign direct investment or bolstering of the start-up economy follows suit.”

A Cultural Reckoning

The Data Bill represents more than policy, it’s a cultural turning point. In 1710, Westminster Parliament passed the Statute of Anne, establishing copyright terms that protected authors’ rights for 14 years. That principle has served the UK’s economy and culture for over 300 years, evolving through numerous updates, including international agreements, and adapting to new technologies while preserving its core spirit.

Now, as Beaumont Nesbitt warns, Parliament risks “strangling the new model for creators at birth.” In an era where an artist’s Name, Image, Likeness, and Voice increasingly drive value, allowing unlicensed use threatens not just revenue but the entire incentive structure for creativity.”

The Stakes Couldn’t Be Higher

UK Music CEO Tom Kiehl described the bill’s passage as a “pyrrhic victory at best.” The music industry has made its position clear: this isn’t just about revenue, but about agency, authorship, and fairness in a rapidly changing technological landscape. The creative sector generates £124 billion annually for the UK economy. The government’s gamble is that AI growth will more than compensate for any damage to traditional creative industries.

Whether the UK can remain both a world leader in culture and a hub for trustworthy AI depends on closing the transparency gap fast. Without action, Britain risks trading a short-term AI lead for the long-term erosion of its most iconic export: creativity. The Data Bill aimed to modernize infrastructure but instead ignited one of the most urgent cultural fights of the AI age.

By rejecting transparency, despite wide support across sectors, the government has given generative AI firms a powerful advantage: access without accountability. The next legislative chapter will define not just the future of British music, but the country’s reputation as a place that values creativity in the age of artificial intelligence.

The question remains: is the UK willing to sacrifice its own cultural industries for a marginal advantage in the global AI race?