The UK Government is currently engaged in a significant public consultation to address the evolving relationship between copyright law and artificial intelligence, the results of which could shape the future of the nation's digital landscape. The Government's stated ambition is to position the UK as a world leader in AI, yet it needs to resolve a fundamental tension: how to balance the rights and needs of content creators and rights holders with AI developers' desire to access significant amounts of high quality data.
Central to this discussion is the complex issue of text and data mining (TDM). Rights holders argue that it is difficult to control the use of their works and obtain appropriate remuneration, while AI developers, conversely, require access to vast quantities of data, and have disputed the extent to which copyright law should restrict this. At the request of the House of Commons Science, Innovation and Technology Committee, Google and OpenAI have now published their responses to the consultation.
The Government has outlined four potential paths in its consultation document:
- maintaining the status quo (Option 0)
- strengthening creator protections (Option 1)
- implementing a broad TDM exception with no opt-out (Option 2)
- adopting a TDM exception with opt-out for rights holders (Option 3)
In its consultation, the Government stated that its preferred approach is Option 3, which would introduce a new TDM exception allowing AI training on copyrighted works, unless rights holders have "reserved their rights" or "opted-out" through agreed mechanisms. This model is intended to align the UK more closely with the European Union's framework and strike a balance to support both creative industries and the AI sector. The Government has sought to reassure rights holders by saying that this opt-out system would be underpinned with greater transparency requirements for AI firms regarding their training data sources. However, in light of significant opposition from across the creative industries, reports have suggested. Many that the Government's stance may be softening in favour of considering the options more broadly. For further information on the detail of the consultation, please see our article UK Government consultation on copyright and AI: A 'win-win'?.
In its response to the consultation, OpenAI has firmly rejected Option 3, calling it "unworkable" and arguing that it does not balance the Government's objectives as set out in the consultation. OpenAI contends that an easy opt-out mechanism for rights holders could render most copyrighted data unavailable for TDM, hindering AI growth and primarily benefiting large companies with existing data reserves. It points to the EU's opt-out framework as demonstrating "significant implementation challenges" due to a lack of clear technical standards, leading to uncertainty. Instead, OpenAI strongly advocates for a broad data mining exception (Option 2) to be introduced, which it believes would provide a clear, predictable regulatory environment necessary to attract investment and talent to the UK. It suggests Option 2 could include requirements of (i) limiting use to lawfully obtained copies, (ii) restricting further distribution, and (iii) implementing measures to prevent infringing outputs.
Google, conversely, indicates support for Option 3, albeit in an amended form (and one which does not necessarily lead to remuneration for rights holders). While acknowledging Option 2 is the "most competitive" for the UK's AI leadership ambitions, it believes that rights holders "should have choice and control". It therefore advocates for TDM for any purpose with a coupled rights reservation, plus a commercial research exception. However, Google asserts that training on the open web "must be free", rejecting the idea that rights reservation automatically implies remuneration rights for rights holders.
Both companies share "significant concerns" regarding what they describe as "excessive transparency requirements" under Option 3. They argue that providing detailed disclosure is unworkable, risks exposing trade secrets, and could hinder AI development and the UK's competitiveness. On liability for infringing outputs, both largely agree that this should fall on the end user, not the model developer, describing such infringing outputs as a "bug" and highlighting developer efforts to reduce them through the use of guardrails. Both suggest that issues such as digital replicas (deepfakes) and synthetic likenesses should be addressed through targeted legislation rather than through copyright law.
Comment
In essence, while both Google and OpenAI recognise the benefit of updated copyright rules to support AI in the UK, they propose different fundamental approaches: OpenAI favours a broad TDM exception (Option 2), while Google supports an amended opt-out (Option 3) built on existing technical controls and the principle of free training access to data on the internet. Both oppose stringent transparency mandates and largely agree on user liability for infringing outputs and addressing digital replicas outside copyright.
The Government has received over 11,500 responses to the consultation and has stated that no final decisions have been made, noting they are carefully considering submissions and continuing to engage with stakeholders.
However, alongside the consultation, the Data (Use and Access) Bill is making its way through Parliament, and has led to significant debates about the question of copyright and AI training in both Houses. With the intention of strengthening transparency surrounding data usage, the House of Lords' recent amendments to the Bill would require developers to disclose to copyright holders when their data has been used in training AI models. Unsurprisingly, however, the Government has said that it will consider the relevant copyright issues within the framework of its consultation.
For updates on the consultation and the various intellectual property cases concerning generative AI currently proceeding through the courts, sign up to our Generative AI Tracker.