UKIPO Consultation on AI and IP: What are the implications for UK copyright law?

The consultation comes seven months after the UKIPO published is response to a “Call for Views” on AI and IP which surveyed a wide range of issues relating to the interaction between AI technology and patents, trade marks, copyright, designs and trade secrets (read our summary of the Government’s response here). Having started with a broad focus in its Call for Views, for this consultation the UKIPO has focused on the patentability of AI technology, the subsistence of copyright in AI generated works and the scope of exceptions for text and data mining (TDM). However, other IP rights are not excluded entirely; the interaction between copyright and design rights is mentioned in the context of AI generated works and possible TDM exceptions to the sui generis database right are mentioned. The consultation is accompanied by an Impact Assessment, which provides further insights into the UKIPO’s thinking.

This article considers the policy options put forward by the UKIPO in relation to computer generated works and TDM. It also looks ahead to where the consultation may lead us.

Computer Generated Works

Computer generated works (CGWs) are those created by a computer and which do not have a human author. They are distinct from works created by a human author using a computer as a tool, which are subject to copyright protection in the usual way. In contrast to most jurisdictions, the UK already offers a form of protection for CGWs by way of section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA). This protection lasts for 50 years from the year in which the work was created, and the author of the work is deemed to be the person who made the arrangements necessary for the creation of the work.

The consultation highlights that the protection currently offered to CGWs has been subject to legal criticisms. In particular, it is not clear how to square a work created without a human author with the originality test which, following the CJEU’s decision in Infopaq (and the line of case law which followed), requires the work to be an author’s intellectual creation. The Impact Assessment suggests that the protection of CGWs needs to balance incentivising and rewarding investment in AI creativity with the costs to third parties who may wish to use protected material, such as increased prices and reduced competition. Interestingly, the Impact Assessment comments that jurisdictions such as the US and EU which do not currently protect CGWs still appear to provide sufficient incentives for the creation of CGWs, without offering them specific copyright protection. The consultation also raises the broader philosophical question of whether offering protection to CGWs could promote these works at the expense of human creations and devalue human creativity.

In this context the consultation seeks views on whether CGWs should be protected at all and if so, how? Three possible policy options are put forward: make no legal change (Option 0); remove protection for CGWs (Option 1); or replace the current protection with a new right of reduced scope/duration (Option 2). For Option 2, the UKIPO suggests that a new right could provide a much shorter term of protection than the current 50-year term offered to CGWs (e.g. 5 years), allowing third parties to benefit from free use of the work much sooner than is currently the case. The owner of the right would be the same as for the current CGW protection (“the person by whom the arrangements necessary for the creation of the work are undertaken”), which is also used to identify the producer of a sound recording and film and the UKIPO considers to be reasonably clear. The new right would sit alongside any other forms of protection which may apply to the work, e.g. copyright protection for elements which do have a human author.

The Impact Assessment clarifies that any legal changes would be forward looking and would not result in any CGWs which are currently protected losing that protection. Nor would they affect the existing entrepreneurial rights protection offered to films, sound recordings or broadcasts, which will continue to be protected irrespective of whether they are created by a human or an AI. It also notes that UK design law contains provisions for identifying the author of computer generated designs which mirror the test applied to CGWs. A review of these provisions does not form part of the current consultation, although the UKIPO welcomes views on the implications of the policy options for CGWs on the design system.

Comment

The choice between Options 0, 1 and 2 is framed by the UKIPO as the balance of incentivising the creation of CGWs with the cost to third parties, including the users of those works and human creators. The UKIPO acknowledges that its challenge in achieving this balance is a lack of evidence on the incentivising effect of protecting CGWs. The Impact Assessment indicates that the UKIPO has been unable to identify whether there are any clear differences in the levels of AI investment attributable to whether jurisdictions currently protect CGWs. The difficulty with this exercise is (i) the rapid pace at which AI technology (and with it the ability to generate useful CGWs) is developing and (ii) extracting meaningful information about the impact of copyright policy from economic data which is the product of many other non-IP related factors. It will therefore be critical for those who wish to see CGWs protected in the UK to present the UKIPO with evidence of how a presence or absence of protection could incentivise or disincentivise their creation.

The consultation also mentions the problem of false attribution, whereby a person falsely claims that a work generated by AI was actually generated by them in order to obtain copyright protection. This problem exists under the current system of CGW protection as a human could claim to be the author of a CGW to obtain copyright protection for their lifetime plus 70 years, rather than the 50 years from the year of creation which applies to a CGW. However, given that CGWs have only come to prominence in the last decade, it’s fair to say that the difference between these two terms has not really been an issue thus far.

It may however become far more important if either Option 1 or 2 is selected, as the difference will become one of full copyright protection vs no protection (Option 1) or full copyright protection vs shorter term protection (Option 2). The incentives to falsely attribute will therefore become much greater. The UKIPO mentions that the Fraud Act 2006 would penalise people who make false representations for gain. However, given that the distinction between a CGW and a human created work depends on the application of complex case law on originality/intellectual creation, it is difficult to see how prosecutions could be brought against those claiming to be authors of works unless it was very clear that they knew that they were not - nor might not be - the author, such that the mens rea required by section 2 of the Fraud Act could be established to the criminal standard.

The position of third parties will also need to be considered in situations where a work they wish to use is not accompanied by an attribution, i.e. how can they distinguish between CGWs (which will only be protected for a short term) and human created works (which will be protected for much longer)? The distinction may not be apparent from the subject matter of the work and a third party would instead need to investigate the manner in which the work had been created in order to determine whether they were allowed to use the work. This situation could give a new importance to section 97 of the CDPA, which provides that damages are not payable for infringing acts where at the time the defendant did not know, and had no reason to believe, that copyright subsisted in the work. A genuine and reasonable belief that a human created work was actually a CGW could therefore potentially engage section 97.

Text and Data Mining

Text and data mining (TDM) is the application of automated computational techniques to data in order to identify patterns, trends and other useful information and is commonly in the development of AI systems. The issue addressed by the consultation is what should happen where such analysis requires (i) the reproduction of material which is protected by copyright and/or (ii) the extraction of data from a database which is protected by the sui generis database right. The UK was for a while leading the pack in Europe by way of section 29A CDPA. Introduced in 2014, the section provides an exception to copyright for the reproduction of works for computational analysis by someone with lawful access to the work. The exception is however limited to computational analysis for the purpose of non-commercial research and only extends to copyright, not database rights. While useful for those in academia (especially as rights holders are prohibited from excluding the exception using contractual terms), the exception does not provide comfort to those engaged in commercial text and data mining who are left to secure licences from rights holders or run the risk of a claim for infringement.

While an early frontrunner, the UK has now been outpaced by the EU and the implementation of the Digital Copyright Directive, which had a transposition date of 7 June 2021. Falling outside of the transitional period following the UK’s departure from the EU, the UK was not required to implement the directive and the Government decided against doing so voluntarily. The UK therefore currently lacks (i) the exception to both copyright and database rights provided under Article 3 for scientific research by research organisations and cultural heritage institutions and (ii) the exception to both rights for all forms of text and data mining (subject to rights holder opt-out) under Article 4.

The consultation identifies that the balance to be struck with TDM exceptions is between reducing the cost of TDM activities and ensuring that any exception does not act as a disincentive for creators to continue creating works and databases. In this context the UKIPO proposes five possible options:

• Option 0: Stick with only the current exception under section 29A CDPA, perhaps with updated guidance on the definition of non-commercial research.
• Option 1: Improve the licensing environment for the use of works and databases for TDM through educational material, model licences or codes of practices which would assist parties to conclude TDM licences. The Impact Assessment contains an intriguing but passing reference to a legislative backstop for codes of conduct, which suggests that under this option the Government could consider legislating in future if voluntary codes of conducts are not followed. It also mentions the potential use of the extended collective licensing framework.
• Option 2: Extend section 29A CDPA to cover commercial scientific research and database rights. This would provide a slightly broader exception than the EU’s Article 3, as it would be defined solely based on the purpose (scientific research) and the beneficiaries would not be limited to research organisations and cultural heritage institutions.
• Option 3: Adopt a TDM exception to copyright and database rights permitting both commercial or non-commercial TDM but with the ability of rights holders to opt-out. This would effectively be modelled on the EU’s Article 4 exception.
• Option 4: Option 3, but without the option for rights holders to opt out of the exception. This option would be the most strongly in favour of AI developers at the expense of being the least favourable to rights holders.

Comment

The UK clearly faces a risk that the EU will come to be perceived as a more favourable jurisdiction for TDM activities and draw investment in AI research and development away from the UK. However, as the responses to last year’s Call for Views indicates, there is a strong rightsholders lobby in the UK who are concerned that extending exceptions could weaken their ability to derive reward from the use of their works. The evidence obtained by way of the consultation will therefore be critical in determining whether the UKIPO takes the view that TDM can be sufficiently facilitated through individual licensing arrangements or whether intervention is required to correct any current market failures.

Although only briefly mentioned by the Impact Assessment, the possibility of utilising some form of collecting licensing framework is an intriguing one. TDM activities often involve collecting works from many different sources which may involve works owned my many thousands of individual rights holders. Without a single point of licensing for these datasets, the task of obtaining licences covering every work in the dataset is insurmountable and AI developers are faced with the choice of either abandoning the project or running the risk that each individual rights holder will not be concerned enough about the use of their work for TDM to take any enforcement action. This is clearly the type of situation where collective licensing could benefit both the individual rightsholders and AI developers.

While not expressly mentioned in the consultation, another angle which could potentially be explored by the UKIPO is whether a new exception could take into account the relationship between the end product of the TDM and the original works which were used. TDM actives in the US are viewed through the lens of a general fair use doctrine (codified at 17 U.S. Code §107), under which one of the factors taken into account is “the effect of the use upon the potential market for or value of the copyrighted work”. Uses of TDM which generate material which could act as a substitute for a work are therefore less likely to fall within fair use than uses of TDM which generate outputs of a different nature. For example, the use of photographs in TDM to create an image classicisation is more likely to be considered fair use than the use of photographs to create an AI system which itself creates images which could compete in the marketplace with those photographs. This assessment of the end result of the TDM process is lacking from the EU’s commercial TDM exception and could go some way to reconciling the interests of rights holders and AI developers.