The Government has launched a much-anticipated major consultation on the relationship between copyright and artificial intelligence, setting out – among other things – proposals on how UK copyright law should apply to the training of AI models.

We have commented on this highly-contested area on multiple occasions (see, for example, here). Actors on both sides of the debate have sought clarity on the status of the law for some time; clarity which has often proved elusive.

By way of background, as early as 2023, in his ‘Pro-innovation Regulation of Technologies Review’ Sir Patrick Vallance recommended that the Government should “announce a clear policy position on the relationship between intellectual property law and generative AI to provide confidence to innovators and investors”. Having abandoned initial plans to amend the text and data mining exception in s.29A of the Copyright, Designs and Patents Act 1988 (CDPA) to allow the use of copyright works to train AI models for commercial purposes, the previous Government sought to respond to Sir Patrick’s recommendations by proposing a code of practice which would be developed by the UKIPO alongside AI developers and representatives of the creative industries. However, this too proved short-lived, and the previous Government announced on 6 February 2024 that its plans for a code of practice on copyright and AI had been abandoned.

Since then, we have largely been left in limbo, waiting to hear the proposals of the new Government. In the meantime, many in the creative industries (and beyond) urged the Government to ensure that their interests as right holders are be ignored or subordinated to the interests of developers of AI models (see, for example, here).

Finally, the wait is over and the Government has come forward with its proposed approach.

The approach notably avoids the words ‘opt-out’, but that is in effect what is being proposed: the data mining exception under the CDPA would be expanded to permit data mining on copyright works for commercial purposes, but right holders can opt out (or, as the consultation puts it, “reserve their rights”), as under the EU DSM Copyright Directive.

According to the consultation document, the proposal is a compromise between the two extremes of either (1) strengthening copyright to the extent of requiring licensing in all cases of data mining (thereby potentially making it harder for AI models to be developed in the UK) or (2) introducing a broad data mining exception which would allow data mining on copyright works irrespective of the right holders’ wishes (thereby meaning that they could not control or seek remuneration for the use of their works). Instead, the proposal “seeks expressly to balance the objectives of access to works by AI developers with control over works by right holders supported by increased trust and transparency for all parties. AI developers would be able to train on material to which they have lawful access, but only to the extent that right holders had not expressly reserved their rights. It would mean that AI developers are able to train on large volumes of web-based material without risk of infringement. Importantly right holders are also able to control the use of their works using effective and accessible technologies and seek payment through licensing agreements”.

Alongside – and underpinning – this proposal would be a requirement that AI developers are more transparent about the sources of their training material. The consultation document explains that, as things stand, “it is difficult for right holders to determine whether their works are being used to train AI models. Some developers use works from databases which are open to the public and can be scrutinised by third parties. But others do not disclose the source of the works they use in training, or only disclose limited information about them. This can make it challenging for right holders to enforce their rights”.

Quite what will be expected of AI developers in terms of greater transparency is not made clear. The consultation explains that regulation may be needed in order that AI firms disclose the use of specific works and datasets, are required to keep records, or provide certain information on request. However, it accepts that given the massive size of these datasets, as well as potential restrictions that may be placed upon their being disclosed (such as underlying commercial contracts or a risk of compromising trade secrets), further work is required to establish the extent and nature of any transparency requirements.

Whilst trying to resolve the tension between copyright protection and training AI models is the central focus of the consultation, it does briefly turn to other important matters. For example, it contains a section on the provision at section 9(3) of the CDPA which provides specific protection for “computer-generated works” (“CGWs”), pointing out that “should consultation reveal insufficient evidence of positive effects from CGWs protection, our preference will be to remove it”. There is also a section that considers the role of output labelling, asking whether generative AI outputs should be labelled as AI-generated. Finally, the consultation addresses the question of how to tackle the growth of deepfakes and other “digital replicas”, and acknowledges the calls from some in the creative industries that US-style ‘personality rights’ should be introduced to the UK (in part as a result of the gaps in existing English law governing deepfakes, as we have discussed here). Interestingly, while the Government is clear that “we are not consulting on specific proposals on personality rights”, it nonetheless states that “we welcome views on whether the current legal framework remains fit for purpose” and that there “will be opportunities to engage with us on this topic in the future”. Watch this space…

The consultation is open until 25 February and can be read in full here.