How does the UK’s intellectual property regime stand up in the face of emerging AI technology? Running from 7 September to 30 November 2020, the UKIPO’s call for views on AI and IP sought to find answers to a wide range of questions covering patents, copyright, trade marks, designs and trade secrets (see our summary of the questions and the background to the Call for Views here). With 92 responses from individual IP attorneys, trade bodies, industry associations, tech sectors, creative industries and other sectors the Government had a lot to consider.
The Government’s highly anticipated response, published on 23 March 2021, summarises the views it received and sets out its conclusions and next steps. The response paints a mixed picture, depending on the IP right in question.
The Call for Views did not elicit a strong push for changes to current trade mark, design or trade secret law to deal with AI technology, and the Government has concluded that the current regime is adequate to deal with AI technology in the short to medium term. No action is therefore planned in relation to these IP rights, save for ensuring the provisions relating to computer-generated designs are aligned with any changes made regarding computer-generated copyright works (see below) and generally monitoring the position as AI technology develops in the medium to long term.
AI generated works
The copyright and related rights questions gave rise to a diverse range of views. As expected, the most hotly contested was the availability of copyright protection for works which are created using AI technology, without input from a human author. Many in the creative industries emphasised the fundamental importance of human creativity in the justification for copyright and expressed concern that granting copyright to works which are mass produced by AI systems without human creative input could devalue and even act as a deterrent to human creativity. They also expressed concern about the lack of clarity regarding the application of the existing provisions for computer-generated works under s9(3) of the Copyright, Designs and Patents Act. Other respondents, especially those from the technology industry, suggested that AI creations should enjoy some form of protection, either under the existing provisions for computer-generated works or through some other related right, more limited in scope than the copyright afforded to human works.
In its conclusion the Government has recognised the points raised by the creative industries. It agrees that the current approach to computer-generated works is unclear, and that there is a case for reconsidering it. It also appreciates the concerns that mass-produced works generated by AI could devalue human creators and agrees that we should not undermine copyright’s central role in rewarding artistic expression and talent. It therefore plans to consult on whether to limit copyright in original works to human creations (including those where AI assisted a human author). This could result in the removal of the existing provisions under s9(3), although the door has not been entirely closed to granting computer-generated works some protection as the Government also intends to consult on whether these works should be replaced with a related right, with scope and duration reflecting investment in such works.
IP Rights in AI training data
While the discussion regarding AI generated works raises issues of fundamental importance to those in the creative industries, a more pressing issue for those developing AI systems is the scope for copyright (and related rights) to restrict access to AI training data. This issue requires a careful balance between the legitimate interests of rights holders to protect and receive remuneration for the use of their works and the economic benefits of AI developers having easy access to datasets in order to foster innovation and reduce potential bias in their systems. Two key dimensions to this issue are the availability of exceptions to copyright for text and data mining activities and the availability of licensing schemes which provide AI developers with a route to access training data.
Given the complexity of the issues involved, the Government concluded that it needs to better understand the ways in which copyright owners currently license works for use with AI and consult on measures to make this easier, including improved licensing or copyright exceptions, to support innovation and research.
AI generated inventions
Whether patents should be granted for inventions which are created by AI systems without the involvement of a human inventor gave rise to mixed views. The current UK legal position following the High Court decision in Thaler v Comptroller General (see our reports here and here) is that patents require a human inventor. Some respondents argued that this position potentially discourages the adoption of AI technology in the innovation process, or could have a detrimental effect on transparency, as applicants would be incentivised to name a human inventor even where the invention was actually devised by an AI. Others expressed concern that if AI inventorship is not permitted, AI generated inventions might not be published. This would, undermine one of the purposes of the patent system: encouraging technical innovation to be shared in exchange for a temporary monopoly over it. Some respondents expressed scepticism that current AI systems could devise inventions on their own and considered there to be no need to change the current position. Some respondents suggested that a legislative provision be introduced in similar terms to that used in copyright for computer-generated works, referring to an inventor as including “a person by whom the arrangements necessary for devising an invention are undertaken”. A sui generis right for AI was also suggested by some, while others expressed the opposite view.
The Government concluded that these issues require further consideration and intends to consult later this year on a range of possible policy options for protecting AI generated inventions which would otherwise not meet inventorship criteria. Legislative change was expressly recognised as a possibility. What the government’s response does not do is clearly identify a range of practical circumstances an AI generated invention would not meet current inventorship criteria beyond the obvious case, such as in Thaler, where an applicant disclaims any human involvement and names only an AI entity as inventor.
Patents for AI
Several questions in the call for views addressed the criteria for the grant of a patent for an invention which involved the use of AI, including inventive step, sufficiency and excluded subject matter. In relation to inventive step and sufficiency there was general agreement that the current law was fit for purpose. For inventive step, a large majority thought it unnecessary to develop any concept of a ‘skilled machine’ to supplement the ‘person skilled in the art’. For sufficiency, although a number of respondents asked whether a deposit system for training data sets or a specification of machine connection weightings was needed to avoid applicants having to include large volumes of information in their patent specifications to make them sufficient. Others reported that such voluminous data should not be needed and thus no consideration of a deposit system was necessary.
The biggest area of debate surrounded excluded subject matter, i.e. whether an invention is a “computer program as such”. Some expressed concern that the current approach reduced incentives for innovation by excluding certain AI inventions from protection. Others said the current balance was the right, in some cases referring to the use of confidentiality protection instead of patent protection. Some cited the UKIPO and EPO’s differing approaches to computer implemented inventions as causing practical burdens on applicants. There was disagreement between respondents as to whether this would require only a change of UKIPO practice or also a change in the law: the underlying law is the same but the EPO and English case law differ, although are intended to lead to the same overall result. A few respondents queried whether this issue should arise, since they took the view that ‘core AI’ inventions were intrinsically patentable as being more than a “computer program as such”. Some respondents here suggested a sui generis right might be better suited to AI, as a fast-moving field, than re-consideration of the scope of the computer implemented inventions exclusion.
As with AI created inventions, the Government concluded that further analysis of these issues is required before changing the law. The UKIPO therefore intends to review its guidelines on patent exclusion practices, in consultation with interested stakeholders, in particular to reduce any lack of harmony with outcomes in the EPO. The government also plans to commission an economic study to enhance its understanding of the role the IP framework plays in incentivising investment in AI, and to use this as an evidence base for judging any further intervention. It will also examine the feasibility, costs and benefits of a deposit system for training data.
The Call for Views comes in the context of the UK’s recently published National Data Strategy, which aims to unlock the power of data for the UK as a driver for economic growth. In light of this ambition, the Government’s future actions will aim to encourage innovation in AI technology and promote its use for the public good, preserve the central role of intellectual property in promoting human creativity and innovation and be based on the best available economic evidence.
Given the pace of development in AI technology and the global competition to attract AI research and investment the Government will clearly have its work cut out to deliver on these objectives. The UK is not alone in reviewing its IP laws to facilitate and attract AI research. Japan, Singapore and the EU have all recently adopted permissive text and data mining exceptions aimed at encouraging AI research and the USPTO has also recently published its own report in AI and IP policy. The EU is also pursuing its own European Data Strategy (read our summary here) which is considering the role of IP laws and other policy mechanisms in facilitating the re-use of data. The UK will therefore have to move swiftly to ensure it quickly homes in on any changes to IP policy required to ensure it remains at the forefront of AI research.