Another change of tack: can a new code crack the TDM riddle?

As part of the continuing debate on how best to balance the rights of AI developers and IP rights holders in the context of text and data mining (“TDM”), the UK Government has recently announced that the UK IPO will be producing a new code of practice in this area. Whilst details are very light at the moment, we understand this code of practice will aim to provide guidance to support AI companies to access copyright protected materials for use in training data, with the Government stating that those AI companies that commit to the code can expect to receive a “reasonable” licence from relevant copyright owners in return. 

Background

Following its consultation on AI and IP, the UK IPO announced, in June 2022, its intention to introduce a new exception for copyright and database rights which would allow TDM for any purpose, with no ability for rights holders to opt out or contract out. That would have significantly broadened the existing TDM exception (which is limited to TDM for the purposes of non-commercial research) and had been largely welcomed by those in the AI sector (see our previous post here). However, earlier this year, following significant backlash from those in the creative industries, various statements were made indicating that the Government would be scrapping its plans to introduce that wider exception (see here). That was later confirmed by George Freeman (then minister for Science, Research and Innovation) at a debate in the House of Commons on 1 February.

In the latest instalment, Sir Patrick Vallance (the Government’s Chief Scientific Adviser and National Technology Adviser) has given his views on the matter in his “Report on Pro-Innovation Regulation of Technologies Review” which was published on 15th March (the “Vallance Report”).

What does the Vallance Report say? 

The Vallance Report highlights various challenges facing innovation and development in emerging digital technologies and puts forward a number of recommendations for the Government.  

With respect to generative AI and TDM, Sir Patrick’s key recommendation is for the Government to (urgently) announce a clear policy position on the relationship between IP law and generative AI. He advocates for a pro-TDM landscape in order to attract investment to the UK and show international leadership, and warns that restricting access to data for training sets would be likely to put the UK at a disadvantage. He also appears to be of the view that enforcement of IP rights should focus more on the outputs of AI systems, rather than the inputs.

At the same time, and to try to balance this headline position with the expectations of those in the creative industries, he also recommends:

• a code of practice and a requirement for images produced by generative AI to be labelled as such (but without further elaborating on these points); and
• that the UK IPO: (i) provides clearer guidance to AI firms as to their legal responsibilities, (ii) coordinates information gathering on systematic copyright infringement by AI; and (iii) encourages the development of AI tools to help enforce IP rights.

How has the Government responded?

In its response, which accepts all of Sir Patrick’s recommendations, the Government has confirmed that it will act quickly to clarify how IP law applies to the AI sector. Most interestingly, it confirms that the UK IPO will produce a code of practice by the end of the Summer. That code will provide guidance to support AI companies to access copyright protected material for use as an input to their models, “whilst ensuring there are protections (e.g. labelling) on generated output” to support IP rights holders. Those AI companies which commit to the code of practice can expect to have a “reasonable licence” offered by rights holders in return, with Adam Williams (UK IPO chief executive) reportedly commenting that “the idea is to make licences more available for people who want them.”

In pulling together the new code, the Government intends to engage with both AI companies and rights holders, in the hope that this will lead to a more balanced code that both sides will agree to and adopt. However, the Government does not rule out the need for new legislation if the code is not adopted or agreed.

Aside from the code, the Government also confirmed that the UK IPO will provide guidance to AI companies as to their legal responsibilities (by the end of Summer), coordinate intelligence on any systematic copyright infringement, and encourage the development of AI tools which assist with enforcement.

Comment

The wave of recent litigation relating to the use of TDM and datasets in the development of generative AI (see our recent blog post and our thoughts in this article from Computer Weekly, to which we contributed) brings all of this into sharp focus and highlights the urgent need to clarify the UK’s position on this issue.

This latest development suggests a move towards one of the other less expansive options raised in the UK IPO’s initial consultation on AI and IP – improving the licensing environment for the purposes of TDM - rather than adopting some form of wider statutory exception (although the Government doesn’t rule out legislation being required in future).

Whilst it’s good to see that the UK Government is still actively engaging on this, given the limited amount of information available at this stage, these latest statements arguably raise more questions than they answer. For example, what will the code contain? What does a “reasonable licence” look like, will a fee be payable? Will this result in an unreasonable administrative burden for the developers? Are we edging closer to a FRAND-style regime for AI applications? We will just have to wait (again) and see. That being said, done properly, and with input from both the AI and creative sectors, producing a code of practice could prove to be an efficient way of reaching a fair middle ground and is consistent with the UK’s adaptive non-statutory basis for regulating AI as set out in the Government’s AI White Paper published on 29 March. 

For now though, the UK remains one of the less friendly places to develop AI as its exceptions to copyright and database right infringement for TDM are narrower than in many other countries (e.g. the EU, US, Singapore and Japan). That clearly does not sit well with the Government’s lofty ambitions with respect to AI innovation, so it seems likely that there will be more to come.

For more information on the risks and opportunities around AI, explore the different publications and podcasts from our Regulating AI series.