The UK AI copyright report published on 18 March 2026 drops the government’s earlier plan for a text and data mining exception with an opt-out and proposes no immediate change to the law on AI training. It does propose removing protection for purely computer-generated works and exploring a new right against digital replicas. For companies that train, license or own content in both the UK and the EU, the practical result is two different regimes to plan for, with no single opt-out that works in both.
Key takeaways
- A broad exception with opt-out “is no longer the government’s preferred way forward”; the government will gather more evidence and keep the issue under review.
- UK law still has no commercial text and data mining exception: the only specific one, section 29A CDPA, covers non-commercial research.
- The government proposes to remove copyright for wholly computer-generated works under section 9(3) CDPA, while keeping protection for AI-assisted works.
- Transparency, labelling and technical standards will be pursued through best practice with industry for now, not legislation.
- In the EU, the opt-out under Article 4 of the DSM Directive and the AI Act obligations already apply, so the two regimes diverge.
What is the UK report on copyright and AI?
The Report on Copyright and Artificial Intelligence was published on 18 March 2026 by the Department for Science, Innovation and Technology, the Department for Culture, Media and Sport and the Intellectual Property Office (IPO), together with an economic impact assessment. Both were required by sections 135 and 136 of the Data (Use and Access) Act 2025, after copyright and AI became a flashpoint during the passage of that Act.
The report follows the consultation that ran from 17 December 2024 to 25 February 2025 and received 11,520 responses. That consultation offered four options: leave the law unchanged, strengthen copyright so that licensing is always required, introduce a broad data mining exception, or introduce an exception with an opt-out and transparency measures. The last was the government’s preferred option at the time, and most respondents rejected it.
What the UK AI copyright report proposes
| Topic | Government position (March 2026) |
|---|---|
| Exception for AI training | Opt-out exception dropped as preferred option; gather evidence and consider alternatives |
| Models trained outside the UK | No change to the law for now; let the courts decide, including the appeal in Getty Images v Stability AI |
| Input transparency | Monitor rules abroad; develop best practice with industry, which may inform future legislation |
| Labelling AI content | Explore best practice with industry; no UK labelling obligation today |
| Technical tools and standards | Keep regulation under review; support market-led standards |
| Licensing | No intervention in the licensing market at this stage; Creative Content Exchange pilot |
| Computer-generated works | Remove the specific protection for works without a human author |
| Digital replicas | Explore options, including a possible new digital replica or personality right |
The government explains that evidence on how copyright affects AI development is “limited and uncertain”, that there is no consensus among stakeholders, and that litigation, EU transparency rules and technical standards have moved since the consultation closed. It will not reform the law “until we are confident” that reform meets its objectives.
Is AI training on copyright works lawful in the UK?
Without a licence, often not. The report itself notes that the relevant exceptions are temporary copying (section 28A of the Copyright, Designs and Patents Act 1988, CDPA) and data mining for non-commercial research (section 29A), both subject to conditions, and that the lack of a broad data mining exception or fair use principle places the UK “among the countries with greater protections for right holders”. In the EU, by contrast, a general exception for text and data mining applies unless the right holder opts out, in machine-readable form for online works (Article 4, Directive (EU) 2019/790).
Models trained abroad raise a separate question. In Getty Images v Stability AI, the High Court held in November 2025 that a model could in principle be an infringing “article” if imported, but that the model in question did not contain copies of the works. The report records that this point is under appeal and proposes no legislation on it meanwhile.
What happens to computer-generated works?
Since 1988, section 9(3) CDPA has protected literary, dramatic, musical and artistic works generated by a computer with no human author. The author is deemed to be the person who made the “arrangements necessary” for the work’s creation, often the person who wrote the prompt, and protection lasts 50 years from creation. The report points out an internal contradiction: these works must be original, but originality is now defined as the author’s own intellectual creation, a human test.
The government proposes to remove this protection, absent evidence of its value, while AI-assisted works with human creative input remain protected, as they are in the EU. AI-generated sound recordings and films could still be protected as such. Until Parliament acts, though, section 9(3) remains in force.
What are the next steps in the UK?
Parliament is pushing in the other direction. On 6 March 2026 the House of Lords Communications and Digital Committee published its report “AI, copyright and the creative industries” (HL Paper 267), and the government’s response was published on 15 May 2026. According to a summary by Berry Smith (8 June 2026), the response confirms there are no current plans to amend copyright legislation, announces a consultation on digital replicas expected in summer 2026 and a working group on whether independent creators and smaller organisations need support to license their content.
In short, the next milestones are consultations and working groups, not a bill. Any reform would then need primary legislation, which leaves time to prepare but no certainty about the outcome.
UK and EU: planning for two regimes
| Issue | United Kingdom | European Union |
|---|---|---|
| Commercial AI training | No specific exception; licence generally needed | Allowed for lawfully accessible works unless the holder opts out |
| Opt-out | No legal opt-out regime; technical signals have no statutory effect | Express reservation, machine-readable for online content |
| Training transparency | No obligation; best practice to be developed | Public summary required from general-purpose AI model providers (AI Act, Article 53) |
| Purely AI-generated works | Protected under s.9(3) CDPA for now; removal proposed | No authorial copyright without a human author’s own intellectual creation |
| Digital replicas | Possible new right under study | Personality and image rights under national law |
What this means for your business
- AI developers: do not assume that an EU-style opt-out check makes UK training lawful. For UK-based training, map licences and the narrow section 29A route.
- Rights holders: keep machine-readable reservations for the EU and, for the UK, rely on licence terms and on monitoring; transparency is not yet mandatory there.
- Owners of AI-generated assets: if UK protection under section 9(3) underpins value, for example in a content library or an acquisition, treat it as at risk and document human creative contributions.
- Licensing: draft UK and EU AI clauses separately, with review clauses that trigger if UK law changes.
- Talent and brands: until a digital replica right exists, protect voice and likeness by contract and through trademarks and passing off where available.
Our team handling AI copyright and digital assets across the UK and the EU can build that dual approach, and META Channel’s regulatory practice covers the EU AI Act within the same group.
When to bring in an adviser
- Before training or fine-tuning in the UK, because the absence of a commercial exception changes the risk profile compared with the EU.
- When a deal values AI-generated content, since the proposed removal of section 9(3) protection affects what is being bought.
- When licensing a catalogue for AI use in both markets, to avoid granting in the UK what you meant to reserve, or the reverse.
- When a model trained abroad is offered in the UK, where the Getty appeal may change the analysis.
Handling the UK and EU positions from one team keeps reservations, licences and public statements consistent; our copyright and digital works team also manages registrations and evidence of authorship.
Frequently asked questions
Has the UK changed its copyright law on AI training?
No. The March 2026 report proposes no immediate reform. The government dropped a broad exception with opt-out as its preferred option and will gather more evidence and consider alternatives. Until any change, AI training on copyright works in the UK generally needs a licence unless a narrow exception, such as non-commercial research under section 29A CDPA, applies.
Are AI-generated works protected by copyright in the UK?
For now, yes in some cases. Section 9(3) CDPA protects computer-generated works without a human author for 50 years, with the person who made the necessary arrangements treated as author. The government proposes to remove this protection, while works created with AI assistance and human creative input would remain protected.
Does an EU text and data mining opt-out work in the UK?
Not as a legal mechanism. The UK has no general data mining exception, so there is no statutory opt-out to exercise. Use of works for commercial AI training in the UK generally needs permission anyway. Machine-readable reservations remain important for the EU and as evidence of your position.
Can IP Global Guard manage AI copyright issues in the UK and the EU together?
Yes. We review your training data, licences and rights reservations against both regimes, prepare AI clauses for UK and EU licences, and coordinate qualified UK practitioners where court action is needed. You keep one point of contact and one strategy for both markets.
How IP Global Guard can help you plan for both regimes
The UK has chosen to wait, and the EU has already legislated. That gap affects how you train models, license content and value AI-generated assets. IP Global Guard, the IP services line of META Channel Corporation Limited, coordinates copyright and AI matters across more than 25 jurisdictions in Europe, Latin America and Africa, with one strategy and one billing relationship.
Tell us where your models are trained and deployed, or which catalogue you want to protect, and in which markets. We will set out what the UK and EU rules require today and what to prepare for. Contact our AI and copyright team.
This article is general information, not legal advice, and reflects the position at the date of publication.
Sources
- GOV.UK (DSIT, DCMS and IPO), Report and impact assessment on copyright and artificial intelligence (18 March 2026)
- GOV.UK, Report on Copyright and Artificial Intelligence, full text (18 March 2026)
- UK Parliament, Government response to the Communications and Digital Committee report “AI, copyright and the creative industries” (15 May 2026)
- Berry Smith, Report on AI, copyright and the creative industries: UK Government response (8 June 2026)
- Directive (EU) 2019/790, Article 4 (text as adopted, 17 April 2019)
- European Commission, AI Act Service Desk, Article 53 of Regulation (EU) 2024/1689
