Text and data mining – A UK Update

At a recent event held by Westminster Forum Projects, Chris Mills of the Intellectual Property Office UK gave an update on the UK’s next steps, now the UK Government has confirmed the end of attempts to shepherd a voluntary code for the use of copyright materials in training AI (see here).

Chris explained:

• The Government does not intend to leave the issue to the market. It plans to set out its policy “soon” and this issue is one of the “top issues” for the UKIPO.
• The issue is not party political (albeit very controversial) so, whichever party forms the next government will want to resolve this in a way that strikes the right balance between two key UK industries (creative and technology) and with a view to promoting economic growth.
• Although there is a clear tension between these industries, there is a blurring of the sides, e.g. because the creative industry will use AI and AI developers want to maintain human creative output as a source of training data.
• The Government is conscious that it has not yet found the “sweet spot” in setting the balance and that any changes to the law, successful or not, would be with us for a long time (given the time it takes to change legislation).
• The Government needs to indicate the broad approach and create incentives for stakeholders to contribute towards developing the detail.
• International harmonization is also desirable. The UKIPO is, for example, speaking with the US Copyright Office. WIPO is an obvious body to seek international consensus. However, the range of views internationally is considerable, so first movers and significant jurisdictions, such as the US and the EU, may “make the weather”, with a de facto standard emerging.

Other issues on the UKIPO’s radar include:

• Transparency requirements for training data and outputs
• Personality rights
• IP rights for AI outputs, which may involve rethinking the purpose of copyright and patents. What do we want to incentivize and how should it be done in a world where, e.g., drug discovery may become increasingly obvious through the use of AI?

Comment

Various stakeholders now expect the UK Government to propose an “opt out” approach to text and data mining similar to that in the EU’s Copyright Directive. They also expect the Government will propose legislation for “transparency” and “credit” in favour of rightholders. Perhaps the transparency and credit requirements will move the UK towards the requirement in the draft EU AI Act for the publication of summaries of the data used to train General Purpose AI. Such a provision could make litigation more likely, as rightholders see written evidence of the use of their content by AI developers.