The UK government has launched a consultation that will remain open until February 2025, seeking input on how the country should address the growing tension between copyright law and the rise of artificial intelligence (AI). The crux of the issue lies in the use of copyrighted works for training AI models. Rights holders, including creators and publishers, argue that AI developers should pay for or obtain permission before using their copyrighted works. On the other side of the debate, AI developers assert that the current legal framework is uncertain, hindering innovation and investment in AI technologies. While legal battles on this issue are ongoing in various jurisdictions, such as Germany and the United States, the UK government has acknowledged the need for clearer legislation.
The consultation puts forward four options for reform. The first option is to maintain the status quo, allowing the courts to continue addressing the issue on a case-by-case basis. However, the government has expressed its dissatisfaction with this approach, considering it insufficient to resolve the current legal uncertainty. The second option suggests an opt-in licensing model, where AI developers would only be able to use copyrighted works for training if they obtain express permission from the rights holders. This model would likely be welcomed by copyright owners but could stifle the UK’s AI sector, which the government is eager to develop. The third option proposes a broad data mining exception, which would allow the use of copyrighted material for AI training without permission, akin to the approach used in Singapore and some parts of the US. This option would be beneficial for AI developers but unpopular with copyright holders.
The final option, a middle ground, would allow data mining but with conditions. Copyright holders could reserve their rights, and developers would be required to be transparent about the materials used to train AI models.
The government has indicated that it views the fourth option as the most balanced, although it faces challenges. For example, questions remain about how the rights reservation model would function in practice, particularly in cases where works are already publicly available or were previously used in training AI models. The issue of retroactive application of any new legislation is also a contentious point, as is the potential impact on non-digital works, such as books scanned into digital formats. Additionally, there are concerns about the transparency of developers, as well as the enforcement of any new measures. A major unanswered question is whether the new laws would apply extraterritorially, given the international nature of the AI industry.
In the end, a potential solution could involve collective licensing agreements between rights holders and AI developers. However, such agreements would require cooperation among rights holders, who have yet to widely embrace AI. This uncertainty leaves the UK in a delicate position, balancing the interests of both rights holders and AI developers while trying to ensure that the country remains an attractive place for AI innovation.
