Running from 29 October 2021 to 7 January 2022, the “Artificial Intelligence and IP: copyright and patents” consultation formed the latest round in an ongoing national conversation between the UK Intellectual Property Office (UKIPO) and interested stakeholders (see here). The consultation sought views on the prospects for facilitating patent and copyright protections for “inventions and creative works which are made by AI”, and for making easier the use of copyright-protected material by AI in innovation and research, through alternative approaches to licensing and text and data mining (TDM) exceptions.
On behalf of the University of Nottingham Commercial Law Centre (UNCLC), we submitted a response to the consultation that addressed the three areas of potential legislative change. Our contribution expressed our views as a research centre with direct interest in the development of IP and commercial laws, and not as an organisation with technical experiences in AI or its applied uses as such.
The existing TDM exceptions under UK copyright law
The current TDM exception is found in section 29A of the UK Copyright Designs and Patents Act 1988, according to which a person is allowed to copy a work for the purposes of carrying out “computational analysis of anything recorded in the work”, provided that it is done “for the sole purpose of non-commercial research”. For the exception to kick in the person must have “lawful access” to the work and any copy must be accompanied by “sufficient acknowledgement”. However, the exception is imperative, in the sense that rightholders cannot contract out of it.
The TDM exception allows the use of automated computational techniques to analyse large amounts of information to identify patterns, trends, and other useful information. The impact assessment document accompanying the consultation acknowledges the importance of TDM for medical and scientific research, business intelligence, data analytics and machine learning. Yet, following the earlier Calls for Views, it emerged that there is room for improvement of this regulatory setting for TDMs. In particular, the impact assessment document stresses that “without government intervention, the cost of using TDM to support AI might be higher, and therefore the growth of UK industries that use TDM might be lower based on lack of access and costs of licensing”.
As a result, the government presented four options for legislative change which could allegedly improve the status quo:
- Option 0: make no legal change and retain the current TDM exception;
- Option 1: improve the licensing environment for using works for the purposes of TDM;
- Option 2: extend the existing TDM exception to cover commercial research and databases;
- Option 3: adopt a TDM exception for any use, with a rights holder opt-out; and
- Option 4: adopt a TDM exception for any purpose, which does not allow rights holders to opt-out.
Considering that many of the respondents in the previous rounds of consultations had called for licensing mechanisms to be improved, the government was seeking views on specific approaches to be adopted in relation to licensing.
While this is a convincing argument, we stressed that the improved licensing system, whatever form it might take, should legitimately question the extent to which intellectual property rights can be transformed from a means of protecting rightsholders to a means of increasing the monetisation potential of their rights, beyond what is already established in their favour in law as fair compensation for uses by others. To this end, we warned that before going into licensing one should take a step back and appreciate that the issue with AI is that it can scan bulks of data faster than humans. Nonetheless, if that access is legitimate and authorised via a licence for a human, there is, in reality, no point in pursuing particular new regulatory solutions regarding AI and TDM; and furthermore, there is no need to look into new exceptions. This shifts the analysis not so much with regard to the scope of the exception or the licensing mechanism, but instead, as Strowel and Ducato argue, towards considering the scope of the reproduction and extraction rights with respect to copyright and database. The same has already been argued by the European Copyright Society, in 2017.
Therefore, we believe that licensing might lead to reinforcing disruptively strong information monopolies in the distribution and use of information and knowledge. The issue with option 1 is that, although it picks a perspective that seems fair in principle, it stands neutral on potential concentrations of licensing power, while exploring the idea of introducing more profit-making. Therefore, while licensing may perform ideally in itself, it cannot entirely guarantee fair market operation.
TDM for any purpose with no opt-out?
We believe that the use of artificial intelligence and machine learning is an opportunity to improve research and not an opportunity for rightsholders to capitalise on uses beyond what their range of rights already provides under the current regime. Therefore, Option 4 seems to confer a more measured approach, since lawful access would still underpin the exception. Then, licences and subscriptions which allow such access would be warranted. A potential combination of improved licensing with a TDM exception for any purpose without the possibility to opt-out could be a viable solution that would promote fairness, but also inspire respect for clearer licensing regimes.
The discussion on TDM has been around for several years now. Considering that the UK is not implementing the CDSM Directive (not for now, at least), which contains two mandatory TDM exceptions, the focus of the UKIPO on revamping its national TDM exception is quite understandable. It goes once again in the direction of incentivising investment in the UK, in a post-Brexit environment, and turning the UK into an AI-superpower.
However, any potential reform must be very carefully approached. In terms of TDM, concerns about licensing are essentially concerns about access: it should not be an issue whether AI will access licensed data and undertake activity with content, if its human user is entitled to do so in the first place.