The UK's AI and Copyright Consultation: the Proposed Outcome

The UK government consulted on copyright and AI between 17 December 2024 and 25 February 2025, and said in the consultation document which option it preferred. A preference expressed in a consultation is not law, and as of this page’s date nothing in UK copyright law has changed as a result of it. That gap is the whole subject here.

What was consulted on

The consultation, run jointly by the Department for Science, Innovation and Technology and the Intellectual Property Office, is published at gov.uk. It set out the problem as one of legal uncertainty holding back both the AI sector and the creative industries, and put four options for text and data mining:

• Do nothing. Leave the existing position, under which the only TDM exception in the Copyright, Designs and Patents Act 1988 is section 29A, limited to non-commercial research.
• Strengthen copyright by requiring licensing in all cases.No broad exception; training requires permission.
• A broad TDM exception with no opt-out. Maximum permissiveness for developers.
• A TDM exception with a rights reservation, plus transparency requirements. The option the government said it preferred.

It matters that the existing UK exception is narrow. Section 29A of the 1988 Act permits copies for computational analysis for non-commercial research where the researcher has lawful access, and it cannot be contracted out of. A 2022 proposal to broaden it to all purposes without an opt-out was announced and then abandoned after opposition, which is the immediate history the 2024 consultation was written against.

The preferred option

The preferred package had three linked parts, and it is a mistake to discuss the first without the other two, because the government presented them as conditional on each other.

1. An exception permitting text and data mining, including for commercial purposes, where the developer has lawful access to the material.
2. A mechanism for rightsholders to reserve their rights,switching the exception off for their material — the same architecture as Article 4 of the EU’s CDSM Directive, described in the Article 4 opt-out page.
3. Transparency obligations on developers about what material was used, on the reasoning that an opt-out nobody can verify compliance with is not a right at all.

The consultation also covered adjacent questions: computer-generated works, which the UK protects under section 9(3) of the 1988 Act in a way most jurisdictions do not and which the consultation asked whether to retain; digital replicas and personality rights; and labelling of AI-generated outputs.

Why it stalled

The response from the creative sector was unusually organised, running through national newspaper campaigns and a widely covered silent album released as a protest, and its central objection was practical rather than philosophical: an opt-out places the burden on individual creators to find and use a reservation mechanism that does not yet exist in standardised form, and material already scraped is not recoverable by opting out afterwards. That is the same unresolved “machine-readable” problem that has dogged the EU provision, arriving in a jurisdiction that had not yet legislated it.

The dispute then moved into Parliament, attaching to the bill that became the Data (Use and Access) Act 2025. Amendments seeking to require transparency from AI developers about copyrighted works used in training were repeatedly passed in the House of Lords and removed in the Commons, producing several rounds of parliamentary ping-pong before the bill completed its passage. The amendments were not carried into the Act in the form their proposers wanted.

What the Data (Use and Access) Act did instead

The Data (Use and Access) Act 2025 received Royal Assent on 19 June 2025. On copyright and AI it did not create an exception, an opt-out or a transparency duty. What it did was impose reporting obligations on the government: to publish an economic impact assessment of the policy options consulted on, and a report on the use of copyright works in the development of AI systems, covering matters including transparency, licensing and access to data, within a period fixed by the Act following Royal Assent. The Act’s text is on legislation.gov.uk.

Read structurally, that is a commitment to produce evidence rather than a commitment to a policy, and it was the compromise that allowed the bill to pass. It is also why anyone waiting for the consultation response has been waiting: the government tied its next move to work it had undertaken to publish.

The position today, and what to watch

What is true regardless of what is announced next: UK law today contains only the narrow section 29A exception, so commercial text and data mining in the UK is not covered by a statutory exception and rests on licensing or on arguments about where the acts took place. The government had not, when this page was written, published a final response adopting any of the four options, and its preference as expressed in the consultation had been publicly qualified in the face of the parliamentary reaction.

Three things are worth watching specifically. Whether any exception, if legislated, is paired with an enforceable transparency duty — without one, the reservation mechanism has no verification path and the objection that defeated the proposal in 2025 returns unchanged. Whether the UK aligns its reservation mechanism with whatever emerges in the EU, since divergent machine-readable signals impose the cost on rightsholders twice. And whether section 9(3) on computer-generated works survives, which is a separate question from training and one on which the UK is an outlier — the contrast with the American position on authorship is set out in the Copyright Office’s copyrightability report.