Simply put
- The UK government has published a report on GenAI and copyright. The report confirms that the previously proposed text and data mining (TDM) exception with an opt-out for rights holders is no longer the government’s preferred option, and there will be no immediate changes to UK copyright law. Instead, the Government will monitor international developments, gather further evidence and engage with industry before making further proposals.
- A number of other key issues remain unresolved, including transparency of inputs and outputs, technical standards and the role of licensing, and the Government supports further stakeholder engagement before reaching a final decision.
- It appears likely that copyright protection for computer-generated works will be removed, but this decision has been deferred pending further scrutiny, and a new consultation on the harms of deepfakes (including the possible introduction of moral rights) is expected this summer.
The UK government has published a report and impact assessment on generative AI and copyright. This report follows on from a consultation due at the end of 2024, which we described in our article, “UK government consultation on copyright and AI: A ‘win-win’?” (Publishing the report was also a legal requirement under the Data (Use and Access) Act 2025).
This report was filed against a backdrop of profound and rapid developments in GenAI models and their impact, accompanied by unprecedented levels of litigation. There are currently nearly 100 lawsuits in the United States, as well as in Europe (including an initial referral to the European Court of Justice) and elsewhere (sign up for our tracker to receive the latest lawsuits and details). There is no doubt that, as a result of this, and also the influence of litigation pressure, both direct licensing agreements and voluntary collective schemes are starting to emerge.
Governments face difficult challenges. We need to meet the UK’s desire to protect the UK’s creative industries while providing a platform for UK investment in AI-driven innovation. So it’s probably not surprising that we decided to spend more time getting these decisions “right.” However, the lack of decisive action, such as further consultations, the establishment of working groups, and monitoring of the international situation, means that important issues remain unresolved, which in itself creates uncertainty (including the possibility of further litigation).
Will the UK introduce a copyright exception for AI training?
No, at least not yet. The main headline of the report is that changes to UK copyright law will not be forthcoming and that the government does not consider introducing a text data mining (TDM) exception with an opt-out for rights holders (as originally stated in the consultation) as a preferred option. Instead, it will gather more evidence about how copyright law is impacting AI development in the UK and engage with industry on issues such as input transparency and technical tools and standards, while keeping an eye on what happens internationally.
The proposed TDM exception should operate in a similar manner to that currently in place in the EU under the Digital Single Market Copyright Directive. However, as many observers have noted, the EU’s opt-out framework presents both technical and legal challenges. It is therefore perhaps not surprising that it is no longer the government’s preferred option, given the gaps in the evidence base. Rights holders will be happy with this, but it’s worth noting that it doesn’t appear to have been actually abandoned. In contrast, the government has more clearly stated its intention not to introduce broad TDM exceptions without rights holder opt-outs at this time. It appears that a TDM exception with an opt-out may still be considered as an option if further evidence that the Government seeks to gather suggests that intervention is needed in parallel with international trends. In fact, other options may emerge in the future, including more focused exceptions, such as exceptions based on the use of content for “science and research.”
The status quo will continue for the time being, but that in itself raises many questions. In particular, much of the current complexity stems from the territorial nature of copyright, with most of the AI models used in the UK being trained elsewhere, usually in the US. In the US, copyright law may be more permissive to the use of copyrighted materials in AI training due to the fair use framework (although this is also not yet resolved). Indeed, the status of overseas-trained AI tools was identified in the Data (Use and Access) Act as a specific topic for governments to consider in this report.
In fact, the most popular option in the government consultation was for a stronger copyright framework that would require a license. all AI development (including models trained elsewhere). The EU aims to take a “market access” approach under EU AI law. If general purpose AI (GPAI) models are brought to market in the EU, they must comply with EU copyright law and its requirements for transparency. Meanwhile, the UK government appears intent on letting the legal battle between Getty Images and Stability AI (which is expected to be heard by the Court of Appeal later this year) come to an end on its own before giving an answer as to what the government intends to do regarding overseas-trained models. As a result, we are still far from resolving this issue, and for now it is up to the UK courts to decide based on copyright laws established in the pre-AI era, which could lead to significant changes in the future, although the report does not indicate how they will change.
What other problems need to be solved?
There are many other important issues on which the government has not been able to make a clear position. It said it will work with industry and other experts to develop best practices, while monitoring international developments and keeping regulatory needs under review.
Regarding input transparency, rights holders are already sending a very clear message that they want greater and more detailed transparency about the use of copyrighted works to train AI systems, and that they believe this should be mandatory, not optional. The report notes that while AI companies are not necessarily opposed to transparency, they would like to see a higher level of transparency and industry-driven efforts.
Again, the EU experience is instructive. Providers of GPAI models are required under EU AI law to provide summaries of their training data in order to access the EU market, and enforcement is expected to begin in August this year (although the report notes that many AI companies have not yet published summaries of their training data).
Regarding output transparency (requiring certain AI-generated works to be labeled as such), this is also subject to further industry engagement, although Liz Kendall, Secretary of State for Science, Innovation and Technology, has said that an interim report on output labeling will be published in the autumn. On the other hand, the development of technological tools and standards is seen primarily as a matter of market advancement, with governments playing a supportive role.
Will the government intervene in AI licensing?
When it comes to licensing, rights holders and AI companies are largely in agreement: Licensing is a commercial negotiation between private parties, and governments should not intervene. Indeed, direct licensing agreements and collective frameworks are beginning to emerge, but as always, there is a risk that some will be left behind. Governments therefore see their role as creating the market conditions that allow licensing to flourish, through transparency of inputs and best practices around technical tools and standards, as described above.
The government is also highlighting the Creative Content Exchange (CCE), announced in last year’s Creative Industries Sector Plan, which aims to be a “trusted marketplace” for trading digitized cultural and creative assets. CCE is already working in a pilot phase with a small number of museums, and expects to have a testing platform up and running by the summer.
Does copyright protection still apply to computer-generated works?
The UK is one of the few jurisdictions that provides for copyright protection in relation to computer-generated works (CGW), although the boundaries of the relevant provisions in relation to AI-generated works have not yet been tested.
During the consultation, the Government indicated its intention to lift this protection unless there is sufficient evidence of its positive effects. Most respondents to this question agreed, especially considering concerns about AI content undermining human creators. According to the report, it appears that protection for CGW will eventually be lifted, but for now the government will once again monitor its use and impact (although it was not immediately clear why a final decision on the matter was postponed).
However, this does not leave any major holes in the legal framework. However, as long as it meets the test of originality, AI-I was saved Copyrighted works, even if generated by AI, will continue to be protected in the same way as entrepreneurial works such as sound recordings (as they do not have an originality requirement). It is also worth noting that the treatment of computer-generated designs is also a subject of government consultations on the framework of design law.
Will governments take action against deepfakes?
Although the issue of deepfakes (or in the more neutral term “digital replicas”) was introduced only as an aside during the consultations, it is clear that the government is concerned about the problematic use of personal likenesses. We therefore want to understand the case for giving individuals, from celebrities and creative professionals to the general public, greater control over how their likeness, voice, and personality can and cannot be digitally replicated, while also recognizing the opportunities that AI innovations may present.
That’s why we will introduce further consultation this summer on how to tackle the harm of deepfakes while protecting legitimate innovation. This includes whether to create new moral rights, raising questions that extend beyond intellectual property law to areas such as data protection and privacy. Individual actors wish to be closely involved in the consultation process.
What is happening in the wider international situation?
The interaction between GenAI and intellectual property rights is a complex issue that lawmakers around the world are grappling with. In the United States, the White House has just released a national policy framework on AI. The U.S. government has said that training AI models on copyrighted material does not violate copyright law, but it acknowledges that there are opposing views and supports that courts can resolve the issue.
In the EU, members of the European Parliament recently passed a non-binding resolution on copyright and generative AI, proposing to the European Commission that it double the extraterritorial application of tools placed on the EU market regardless of where they were trained, create a new licensing market including sector-specific voluntary collective licenses, and give the EUIPO a central role in managing licenses and opt-outs. All of this needs to be seen in the wider context that the Digital Single Market Directive itself is due for review this summer, with the EU potentially reconsidering its own TDM framework.
conclusion
With the publication of the government report, we are still in something of a holding pattern, and it seems unlikely that answers will emerge from the lawsuit any time soon. Discussions will continue, with further consultations, stakeholder engagement, court decisions and commercial licensing arrangements expected to develop in the coming months. Shaping the future AI/IP environment requires active and early involvement in these processes.
