Artificial Intelligence and Intellectual Property – the Government’s Plans to amend UK law – or not!
On 28 June 2022, the UK Government published its updated conclusions to the consultation it launched in October 2021 which looked at how patents and copyright, which respectively are intended to reward inventors and creators by protecting inventions and creative works, might usefully be adapted to encourage innovation and creativity in the use of Artificial Intelligence (AI). The updated conclusions can be found here.
The Government had sought evidence on a number of options as to the way AI could interact with patent and copyright laws, recognising that AI is playing an increasingly important role in technical innovation and artistic creativity – recent examples include DeepMind’s AlphaFold 2 which has effectively solved a long standing and important problem in computational biology of predicting three-dimensional protein structures, and OpenAI’s Dall-E2 which can create artwork based on a natural language description of a scene.
The Government focused on three key areas:
- Whether any changes were required for copyright protection for computer-generated works (CGWs) – i.e. works made without a human author.
- How to deal with text and data mining (TDM), which is critical, indeed arguably where it can add most value, to much AI use and development, in the context of copyright – whether by creating fair dealing type exceptions or by some form of compulsory licensing.
- How to deal with patent protection for AI-devised inventions.
For computer-generated works, the Government does not intend to change the law. CGWs are protected under UK law - somewhat unusually compared to the laws of many other countries - by the grant of copyright to, inter alia, “original literary, dramatic, musical or artistic works” without reference to their creation by a human. This is confirmed by the authorship provisions in Section 9(3) of the Copyright, Designs and Patents Act 1988 where the Act recognises that the “author” of a “computer-generated work” (CGW) is “the person by whom the arrangements necessary for the creation of the work are undertaken”.
The consultation process produced no evidence that this current protection provided by copyright law for CGWs is harmful. Most of the concerns raised by respondents related to the potential of false claims by humans that they, rather than AI, created the work (so extending the period of copyright available for 50 years to the life of the author plus 70 years). The Government considered that existing laws (for example, fraud) were sufficient to deal with this risk, while at the same time acknowledging that the use of AI is still in very early stages. Two points were said to follow: first that a proper evaluation of this question is difficult and second that any change to the law carries the risk of unintended consequences. The Government has indicated it will keep the law under review and is prepared to change it if there is evidence to support a change.
Text and Data Mining
Text and data mining means using computers to analyse often vast amounts of data to identify patterns, trends and other useful information. Such analysis could not be done sensibly or within a reasonable time frame by humans. TDM is critical for training and testing AI systems, and also has many other uses, for example in academic and other research, and in the fields of journalism, marketing and business analytics. The Government recognised that, whilst data and the trends and concepts that might be derived from its analysis are not protected by copyright, the source of the data is often embedded in a work protected by copyright or by the database right.
The extraction and reproduction of such data – an integral part of the process of analysing it using a computer technique – will often result in an infringement of copyright in the underlying work. Some rights owners are prepared to grant licences to their works to permit TDM but many do not and the process of obtaining licences results in considerable cost, both legal and from having to pay royalties. This can create a good deal of friction for researchers and developers of AI systems, sometimes to the detriment to the creation of tools likely to be useful for public health, society and economic endeavour. Many rights owners argued for no change to the current system in case that change results in loss of revenue and control over their works.
Possibly surprisingly, the Government grasped the nettle and it has indicated it will introduce a new copyright and database right exception which allows TDM for any purpose, supplementing the current position which allows TDM for non-commercial purposes. Thus, rights owners will not be able to charge for UK licences for TDM and will not be able to contract or opt-out of the exception. Rights owners will have safeguards to protect their content, principally by the legal requirement that access to the underlying data must be lawful (probably similar to the existing qualifications set out in the provisions in Section 50 of the Copyright, Designs and Patents Act 1988). A rights owner will be able to choose the platform from which they make their works available, including charging for access via subscription or single charge. They will also be able to take measures to ensure the integrity and security of their systems – although the scope of this reservation remains to be seen.
This brings UK law into line with similar laws in some member states of the EU, and in Singapore and Japan. Note that the exception will not impact other laws such as those which protect personal data.
Patent Protection for AI- devised inventions
The Government is not planning to make any change to UK patent law for now. The consultation did not produce any evidence that UK patent law does not sufficiently protect inventions made with AI system assistance. It noted the risks in changing the law, particularly with respect to the right of British patentees to make international patent filings in important markets such as the EU and the US. There was a recognition that this is something that needs to be harmonised internationally. In addition, there was little consensus as to who should own patents for inventions “made by AI” and many respondents felt that AI is not yet advanced enough to invent without human intervention. The question of AI inventorship will rumble on and some will feel this is a missed opportunity to address the issue head-on. The Government will keep this area of law under review with the intention of ensuring that the UK patent system supports AI innovation and the use of AI in the UK.
It might be argued that the Government’s efforts have produced more heat than light given they have indicated there is presently no need to change UK patent and copyright law relating to the ownership and protection of computer generated works and inventions. But this would underplay the importance of the exception proposed to facilitate text and data mining. At EIP, we have advised on a number of economically and societally worthy initiatives in the AI and machine learning areas, the funding and execution of which could be impacted by concerns regarding the uncertainty or increased costs in relation to the position under copyright law. The proposed changes will largely remove these concerns. So, all in all, this initiative seems to have produced a good result and should help to support the upsurge in research and development work in AI and related fields.
If you are interested in finding out more about this topic, please get in touch with your usual contact at EIP or one of the following: Heather McCann, Matt Lawman, Mark Lubbock, David Brinck, Ben Maling.