As widely anticipated, the Government has confirmed, in its response to the AI White Paper consultation, that it has not been possible for the UKIPO to broker a voluntary code of practice between AI developers and rights holders relating to the use of copyright protected materials in the training of AI tools. Discussions on a possible code began in June 2023 but, whilst the Government has described these conversations as providing a valuable forum for stakeholders to share their views, "it is now clear that the working group will not be able to agree an effective voluntary code".
This is not surprising given how far apart the interested parties are – with rights holders concerned at large-scale use of copyright protected content for training AI models (and the lack of transparency in relation to the datasets used for such training), whilst AI developers emphasise that they need access to a wide range of high-quality datasets in order to develop and train AI models effectively to harness the opportunities presented.
Whilst the report confirms that the Code of Practice discussions have not been fruitful, it is fairly vague on next steps: we are told that DSIT and DCMS ministers will now lead a period of engagement with AI developers and rights holders, with the aim of ensuring the "workability and effectiveness of an approach that allows the AI and creative sectors to grow together in partnership." In particular, there will be exploration of mechanisms that will provide greater transparency for rights holders to understand whether their content has been used as an input. There will also need to be close engagement at the international level on these issues.
The Government's response comes mere days after the House of Lords Communications and Digital Committee published its report following its inquiry into large language models (LLMs) and generative AI. The Committee's report makes a number of key recommendations generally in relation to the Government's approach on AI, including in relation to copyright. Again, the evidence before the inquiry made very clear the stark nature of the dispute. Rights holders such as the Publishers' Association submitted that LLMs are infringing copyright on "an absolutely massive scale", whereas Meta argued that conducting text and data mining on publicly available and legal accessed works should not require a licence, as it did not amount to copyright infringement, and OpenAI stated its view that "copyright law does not forbid training".
Noting the complexity of the issue (including around identifying relevant acts of infringement, as well as potential defences), the report chastises the Government in its evidence for suggesting that the answer was to wait for the Courts to provide clarity – not least, given that (as noted by Meta's representative), such cases would take "a decade or more" to work through the system, and cases may be decided on narrow grounds or settle.
Ultimately, having heard evidence on all sides of the debate, the Committee came down clearly in favour of rights holders in its report: in its view, the current copyright framework is failing in its objectives of rewarding creators for their efforts and preventing others from using their works without permission. It concludes "…the Government has a duty to act. It cannot sit on its hands for the next decade until sufficient case law has emerged". Instead, it calls upon the Government to publish its view on whether copyright law provides sufficient protection and, if there is major uncertainty, set out options to update copyright legislation.
The report goes on to make recommendations in relation to the then proposed Code of Practice which has now fallen by the wayside. Identifying a middle ground as a baseline for developing a Code of Practice was always going to be extremely difficult. Now, it remains to be seen whether the further engagement to come led by DSIT and DCMS will likewise struggle to reach a mutually acceptable solution. In the meantime, we await the Government's further proposals on the way forward on this complex issue (and the risk that this issue may simply be left for a future Government to resolve).
Related coverage
World Intellectual Property Review