The UK government has published its latest consultation on AI and copyright. This will explore several aspects of AI and copyright and is likely to become a highly contentious issue in the coming months, with a clear clash between the interests of the UK’s creative industries and the government’s aim of fostering greater tech investment.
Background
I said “latest consultation” because if you are getting a sense of déjà vu it’s because we’ve been here before. In 2021, the previous government already conducted a similar consultation which understandably did not get as much attention as this one, but the emphasis as the time was to analyse the role of computer generated works (CGW) in UK copyright law present in s9(3) CDPA, with a look at the text and data mining (TDM) existing exception as a secondary issue. The government published its response to the consultation in June 2022, and it contained two main findings related to copyright, the first one was that there would not be a change to the existing CGW provision, and the second was that the government proposed a change to the existing TDM exception present in Art 29A CDPA to a much broader one, in fact it would have been arguably the broadest exception of its kind in the world. The consultation however did not lead to new legislation being proposed, in part due to opposition from various sectors, but also because the government knew that it had its days counted, and did not consider this something that would require a long political struggle.
As the government did not have the stomach to try to push any new legislation, it tried to make a compromise in 2023 by trying to bring the tech industry and the creative industries together to try to negotiate a code of conduct, but this effort wasn’t fruitful, and the negotiations ended with no agreement between the parties.
Fast forward a few months, and the UK had an election and the new Labour government came to power on July 2024. I’ve written before about some of the politics of AI, but suffice it to say that the new government felt that it had to do something in this area, so now it has published its consultation.
But why did the government feel like it had to act? As many things that are wrong with the UK, the answer is Brexit. In 2019 the EU passed the Digital Single Market Directive (Directive 2019/790), which contains several changes to EU copyright law. Articles 3 and 4 of the Directive created exceptions for text and data mining. Article 3 establishes a mandatory exception to copyright for TDM by research organisations and cultural heritage institutions, provided they have lawful access to the content. Article 4 extends TDM rights to commercial entities, allowing them to mine any publicly available or lawfully accessed content unless rights holders explicitly opt-out through a reservation of rights, such as through machine-readable metadata.
As the UK left the EU after the Brexit vote, it never implemented these exceptions, which opened up a disparity as the UK still has in place an exception for scientific text and data mining that was passed in 2014, and it is contained in s29A CDPA. It is this disparity what has prompted the current government’s action, as it is felt that it creates a comparative disadvantage with the EU’s more lenient exception. To complicate matters further, the EU has been doubling-down on the opt-out approach, the AI Act has entrenched the opt-out regime by repeatedly referencing Art 4 and the need for AI trainers to respect reservation of rights. Furthermore, the latest draft of the General-Purpose AI Code of Practice, which covers AI training practices contemplated in Art 51 of the AI Act, contains a detailed number of provisions and clarifications on how GPAI models should comply with opt-outs. So at least for the foreseeable future, the EU’s opt-out regime is here to stay.
All of the above is what has prompted the current consultation. But what are the options presented by the government?
The options on the table
Just as the 2021 consultation, the current version deals with two main legal questions: what to do with the existing TDM exception under s29A CDPA, and what to do with the provision on copyright for computer generated works contained in s9(3) CDPA, but as stated already, the main emphasis and most of the attention is being directed at the TDM exception.
The consultation has the following options for the TDM regime:
Option 1: Strengthen copyright requiring licensing in all cases. This is likely to be a popular option with copyright industries, and it would require anyone training a model in the UK to have a licence to do so. This option would also cover models trained outside of the UK, even if they were created within a more lenient enforcement system. This appears to be the government’s least favourite option though, as they have stated in the consultation paper that this option is “highly likely to make the UK significantly less competitive compared to other jurisdictions – such as the EU and US – which do not have such restrictive laws.” The government has also commented that this option is unlikely to produce more licensing fees. Another possible result from choosing this option would be a reduction of models offered to the UK market, something that may actually be likely to also be popular with people who oppose AI development.
Option 2: A broad data mining exception. This option is pretty much the same as the government’s proposal from the 2022 consultation response, and it would create an extremely broad exception to copyright, going even further than most other regimes, allowing AI training for even commercial purposes even without a reservation of rights. The previous government justified this option as something that would make the UK more competitive as it would offer more to AI companies than the EU’s exception. I think that this option is the least popular of all by far, and I can’t see even the AI industry advocating for something that is so broad. The government rightly comments that this would severely hinder the growth of the creative industries, so I cannot see this being adopted at all.
Option 3: A data mining exception which allows right holders to reserve their rights, underpinned by supporting measures on transparency. This is the government’s preferred option, and it would generally bring the UK in line with the EU’s practice, although the proposal would also have a built-in transparency requirement akin to the one that is also present in the AI Act. This is presented as the “balance” option, it allows AI firms to train with more certainty, and also gives copyright holders the option to reserve their rights. This option would require robust transparency measures, as well as the creation of opt-out standards, as well as fostering collective licensing deals. The consultation paper explains that by requiring an opt-out, this would make sure that their works can only be used for AI training with their permission, likely though licensing deals. What is not said, but should be understood, is that any firm not complying with the reservation of rights would be liable for copyright infringement.
The other main copyright question is what to do with the existing provision in s9(3) CDPA, which stipulates that the author of a work that has been generated by a computer without human intervention is the person who made it possible for the work to be created. This has been the subject of some debate, and considered to be a bit of an oddity internationally, although I am on record as thinking that this is a pretty good approach to CGWs, but I will completely agree that I am in a minority (more of my thoughts on this subject here). While the previous consultation concluded that there should not be a change in the existing provision, the present consultation proposes the following options:
Option 0: No legal change, maintain the current provisions. As stated, this was the previous position, but the paper is strongly averse to this. In short, an overarching objective of the consultation is to try to bring the UK in line with other countries, and in this area we are the outliers, so doing nothing is not really an option.
Option 1: Reform the current protection to clarify its scope. This would be likely to keep the CGW provision, but amend it to clarify it in line with existing copyright originality requirements. This option would leave the door open to copyrightability of these works, but they would have to meet the standard of being the author’s own intellectual creation, so some sort of human intervention would be needed. The government would only choose this if it can find evidence that CGWs have economic value.
Option 2: Remove specific protection for computer-generated works. This is the government’s preferred option, as there is no evidence that the presence of copyright protection for outputs has economic benefit, and on the contrary, there is strong opposition from some creators.
There are other issues contained in the consultation that I will not cover, such as whether the copyright infringement regime needs to be changed to consider infringing outputs, the interaction of AI and designs, licensing issues, the use of AI in education, how to treat models trained in other jurisdictions, and whether the transient copy exception needs to be revisited (I discuss a bit about that exception here).
What next?
This is likely to be a consultation that will produce a good amount of responses from various sectors (I intend to submit my own response soon), after the consultation is closed we can expect a government response within a few months, my guess would be sometime in mid-2025. The main issue of contention is likely to be the TDM exception, and there is already a considerable amount of pushback to the government’s preferred option online, so it will be interesting to see if that is translated into consultation responses.
However, consultations of this nature aren’t a popular vote, and the government is likely to take into account the opinion from various sectors when making its decision. The consultation paper is written in a way in which there can be little doubt what its preferences are. In the end, this will come down to a political decision and making a call of which interests to put first. In the end the government seems to be betting hard on AI investment, and a perceived competitive disadvantage with other countries, particularly the EU, could help to seal the deal. There is also the looming prospect of an extremely pro-AI technocratic Trump administration, which would further entrench fears of the UK falling behind. On the other side of the debate, the copyright industries are likely to make the case that opening the existing exception will be detrimental to the UK’s influential creative industries. Will the considerable pushback be enough to change the government’s preferred option? Only time will tell.
I do believe that whatever happens with the TDM exception, the days of s9(3) and the CGW provision are numbered.
Concluding
This is an exciting time for those of us interested in AI and copyright. The consultation is likely to offer valuable insights into the competing interests at play in this hotly debated topic and whether a compromise is even possible. One thing is clear, though: this consultation is set to dominate the copyright conversation throughout 2025 in the UK, potentially alongside the looming Getty Images case.
Will you be submitting a response?
0 Comments