Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
The UK government has dropped a proposed copyright exception for AI training as its preferred policy, after strong opposition from creative-industry respondents. But that is not the same as banning AI training on copyrighted works or adopting mandatory licensing. The government has not chosen a replacement, so the rules remain unsettled.
On March 18, 2026, ministers said the opt-out approach was “no longer the government’s preferred way forward.” In its latest located update, dated June 29, the government said further work was continuing on creator control and related issues, without announcing a final policy. The government’s report and June update make the distinction important: the proposal lost preferred status, but the policy debate did not end.
What the UK government changed
The proposal at issue was Option 3 in the government’s consultation on copyright and artificial intelligence. It would have created a broad exception to copyright for text and data mining, including commercial AI training, when developers had lawfully accessed the works. Rights holders could reserve their rights, with machine-readable signals and transparency measures intended to help developers identify material they should not use.
Recommended Free Tools
In practical terms, that would have made use the default: a developer could train on a song, photograph, book or illustration unless the relevant rights holder had reserved the rights in a way the system recognised. That differs from a licensing-first model, where permission would generally be secured before the relevant copying or use.
#1 Best Overall
The government’s March 18 report said Option 3 was “no longer the government’s preferred way forward.” Ministers cited strong creative-sector opposition, uncertainty about economic effects, doubts about whether technical reservations would work in practice, and changing conditions in AI markets and international policy. The statement to Parliament said the proposal had been rejected by the vast majority of the creative industries. That describes the government’s assessment of consultation and engagement; it does not mean every artist or organisation took the same position. Read the ministerial statement.
Status in brief: the opt-out exception is no longer the government’s preferred option; no replacement has been selected; and the announcement did not itself create a new licensing rule or a universal right to stop AI training.
Why creators objected to an opt-out system
The central objection was about who would carry the burden. Under an opt-out arrangement, creators might have to find out that their work was being collected, understand how to reserve rights, apply the signal across sites or platforms, and rely on developers to detect and honour it. Those steps can be particularly difficult for an individual creator whose work appears in many places or has been redistributed without their involvement.
Creative-sector respondents also raised concerns that:
Rank #2
- Works could be used for training without the creator’s permission or payment.
- Systems trained on creative work could generate competing material, affecting the value of that work or the market for it.
- Creators may have limited leverage if licensing becomes the exception rather than the starting point.
- A reservation made after a work had been copied or added to a dataset might not undo that earlier use or reverse training already completed.
- Without meaningful information about which works were used and by whom, rights holders might not know whether a reservation had been ignored or how to pursue a claim.
These are concerns recorded in the government’s report, not a finding that every predicted harm will occur in every case. Nor is “the creative sector” a single bloc: consultation participants included creators and rights holders across music, writing, publishing, visual art, acting, photography and news media, while technology companies, researchers and other respondents also argued for exceptions to support AI development and research.
Organisations including UK Music, the Musicians’ Union, the Society of Authors and the News Media Association were among those reported as opposing or raising concerns about the proposal. UK Music’s consultation response sets out music-industry arguments. The government attributed its change in direction to consultation feedback, creator engagement, gaps in the evidence and the evolving international context; the available evidence does not support saying that a particular artist or celebrity objection alone caused the reversal.
What the consultation numbers do—and do not—show
The consultation ran from December 17, 2024, to February 25, 2025. In the government’s progress report, support among respondents for the options included:
| Option | Reported support |
|---|---|
| No change to copyright law | 7% |
| Broad exception with rights reservation (the preferred opt-out proposal) | 3% |
| Broad exception without rights reservation | 0.5% |
These are figures from consultation responses, not a representative poll of the UK population. The government reported receiving more than 11,500 responses and noted that some were template responses. The percentages therefore should not be read as a national vote or as a precise measure of public opinion. They do, however, sit alongside the government’s qualitative finding of strong opposition among creative-industry respondents. The progress report explains the consultation and its results.
Why “opt out” is more complicated than a button
A rights reservation is a legal or policy mechanism; making it work across the AI supply chain is a separate technical and enforcement challenge. A system would need answers to questions such as: what format counts as a valid reservation; whether it follows a work when a copy is moved or altered; how a developer handles conflicting signals; and how a creator can reserve rights on a platform they do not control.
Timing matters, too. A signal published after a crawler has collected a work may not reach copies already stored in a dataset. Even if a compliant developer stops future collection, that does not automatically delete earlier copies or remove the influence of material from a model already trained. An opt-out should not be mistaken for a guaranteed deletion or model-untraining right.
Transparency is the link between a formal right and an enforceable one. If a rights holder cannot discover which developer used a work, or cannot obtain information sufficient to investigate, neither a reservation nor a licensing entitlement is easy to exercise. The government acknowledged uncertainty about whether technical measures could make an opt-out effective and pointed to questions about how rights reservations operate in the EU, where a comparable approach is part of the text-and-data-mining framework. It did not conclude that the UK had simply adopted that system or that its practical effectiveness was settled. The consultation document discusses rights reservations and technical measures.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesThe four policy paths under consideration
The government assessed four broad directions. Its decision to stop preferring Option 3 did not amount to choosing Option 1.
Rank #4
| Policy path | What it would mean | Main trade-off |
|---|---|---|
| Option 0: Keep the current law | No new broad AI-training exception; existing copyright rules and exceptions continue to apply. | Avoids locking in a flawed new system, but leaves legal questions and enforcement difficulties unresolved. |
| Option 1: Require licensing | Strengthen the rules so that relevant copies made during AI development require permission or a licence. | Could preserve creator control and support payment, but may raise costs and administrative burdens, especially for smaller developers and research. |
| Option 2: Broad exception without an opt-out | Permit broad AI-related text and data mining without a rights-reservation mechanism. | Could reduce transaction costs for developers, but would provide less control to rights holders than the opt-out proposal. |
| Option 3: Broad exception with an opt-out | Permit use by default, with rights holders able to reserve their rights through signals intended to be machine-readable. | Offers developers greater certainty if signals work, but puts much of the practical burden on creators and depends on reliable compliance. |
The government’s impact assessment sets out the options and their expected effects. They involve competing claims rather than a simple choice between “protecting artists” and “supporting innovation.” Licensing may create a route to permission and remuneration, but a workable market needs rights information, records, attribution and enforcement; it may also be easier for large rights holders to access than for individuals. A broad exception can reduce transaction costs, but risks shifting leverage away from creators. Keeping the law unchanged avoids a premature rule while prolonging uncertainty.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What changes for creators and AI companies now?
The proposed new commercial-training exception has not taken effect, and the government has not announced a universal licensing mandate. The March announcement does not automatically stop every AI-related use of copyrighted work, nor does it tell creators to file an opt-out with the government: no universal government opt-out portal is identified in the cited materials.
For now, the application of existing copyright law remains important. Whether a particular AI-related activity involves a relevant copy, falls within an existing exception or requires permission can depend on the facts and applicable law. The government has said it will examine how copyright rules affect AI development and deployment, including cases where a system is developed outside the UK but used within it. The policy debate does not itself decide individual disputes or settle every question about training data, intermediate copies and outputs.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →For developers, the change means the government has withdrawn its preference for a broad opt-out exception, not that it has established a final alternative. Legal certainty, access to training material, and the ability to show what material was used remain live issues.
Best Value
The next test: transparency and workable creator control
In its March statement, the government set out further work on creator control, digital replicas, AI-generated-content labelling and support for smaller creative organisations. The June 29 update confirmed that work was still in progress and did not announce a final replacement policy. The government has referred to reviewing standards, technical solutions and best practice; that is not the same as providing creators with a universal control tool today.
The House of Lords Communications and Digital Committee has urged the government to protect incentives to license and establish robust transparency obligations for large AI developers. Its position reflects a practical point: even strong rights can be difficult to exercise when developers’ training sources are opaque. A useful framework would need to let rights holders discover relevant uses and give them a realistic way to negotiate, reserve rights or seek a remedy. The committee’s response sets out its recommendations.
Several details remain unresolved: whether the eventual approach will be licensing-first, a narrower exception, transparency duties without a new exception, voluntary standards or a hybrid; how smaller creators could participate; what happens to works already used; and how UK rules will address models trained abroad and deployed in the UK. The government’s international comparisons include the EU, United States, Australia and India, but the UK has neither simply adopted the EU approach nor settled the consequences of other jurisdictions’ rules.
Free tools Windows power users keep installed
One-click scans. No signup required.
Digital replicas are related but not identical to copyright in a song, photograph or book. A person’s voice, face, name or likeness can raise questions involving performers’ rights, privacy, contract or passing off, among others. Those questions do not all fit neatly within copyright, which is why the government listed digital replicas as a separate workstream.
The government has reset its preferred approach, not finished the policy debate. The central question now is whether the replacement gives creators meaningful control and a workable path to payment, while giving developers clear rules—and whether transparency makes either goal enforceable.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
