What’s mined is yours – A UK Government consultation on copyright and AI

While it is important to foster innovations in the AI sector, the extent to which AI developers and firms can mine data to train their AI models may concern rightsholders.

The government has recently opened a consultation on this issue, outlining several proposals to encourage further investment in AI without jeopardising the protective measures available for rightsholders.

A Conflict of interests

According to the International Trade Administration, the UK is the third-largest AI market behind the US and China, although several other Asian states (such as Japan and South Korea) are among the fastest growing markets. To maintain the UK’s competitive global position, the government has acknowledged that AI developers must be able to access sufficient training data. For instance, over 75% of respondents to a 2023 UK sector study reported that a lack of training data significantly affected their ability to meet business goals. However, protecting creators under copyright law is equally important to sustaining the UK’s world-leading creative industries, which generated £126bn in gross value to the economy in 2022.

AI training requires mining significant quantities of data. Inevitably, text and data mining models risk infringing copyright laws if any copyrighted works are accessed by the AI model, since the key exception from copyright infringement under UK law only applies if the works are mined for non-commercial purposes.

Rightsholders have expressed concerns about their works being used for AI training without their consent. The government has also noted that creative industries have criticised AI developers for a lack of transparency regarding how their works are accessed and used. This ongoing tension between AI developers and rightsholders is exemplified by the ongoing legal battle between Getty Images and Stability AI, where Getty Images alleges that Stability AI infringed its copyrighted works by using its content to train an AI model without permission.

Striking a balance

In this consultation, the government outlines several proposals to drive growth in the AI sector while maintaining a sufficient degree of protection for creative industries, including:

• An exception to copyright law for text and data mining. This would apply to text and data mining for any purpose, including commercial purposes, but only if the data has been lawfully accessed and the rightsholder has not reserved their rights for the work. Rightsholders could essentially opt-out by reserving their rights under the exception, enabling them to continue licensing their works to be used for AI training.
• Introducing robust transparency measures which may require AI developers to disclose how specific works are used, provide certain information on request, or keep records as evidence that they have complied with the relevant rules.

Potential implications

The AI sector will benefit from an exception which should provide developers with greater access to data for AI training, accelerating growth and innovation in an increasingly significant market. AI firms may also have to accept the added administrative burden and potential compliance costs of ensuring adequate transparency measures are implemented and followed.

For creators and rightsholders, the transparency measures would provide clarity on how their works have been used. Additionally, rightsholders may use paywalls or subscription models to seek payment from AI developers who wish to use their works. However, rightsholders could face a cost burden in complying with the rights reservation.

Currently, the most popular rights reservation standard used by rightsholders is the robots.txt standard. It enables rightsholders to deny access to generative AI web-crawlers and bots, but it does not provide them with granular control over individual works. This means that rightsholders are not able to allow their works to be used for one particular purpose but not for others (e.g. a creator may want their works to be used for AI language training but not for generative AI). The government suggests that machine-readable rights reservation formats should be standardised as far as possible, which should help creators reserve their rights more easily.

The Birketts view

This consultation demonstrates that the government recognises the need to ease tensions between the AI sector and creative industries, both of which are globally competitive and crucial to the UK’s economic growth. Providing an opportunity for stakeholders in both sectors to comment on the proposals is an important step towards reducing these tensions. Industry stakeholders can participate in the consultation until 25 February 2025.