The future regulation of AI: UKIPO consults on patent protection

The UK Intellectual Property Office (UKIPO) is conducting a further consultation on artificial intelligence (AI) and intellectual property (IP) that focusses on copyright and patent issues highlighted in the responses to its previous consultation.

These consultations have been launched against the backdrop of the UK government's aim to position the UK as a "global AI superpower" and the announcement of its National AI Strategy. As well as putting forward proposals on how to manage patent protection for AI-devised inventions, the consultation also addresses the future of copyright protection for AI-generated works and IP issues arising from text and data mining.

The current position

An inventor for the purposes of the Patents Act 1977 must be a human. The Court of Appeal recently held that UK patent applicants must name a human inventor(s) in order to satisfy the statutory requirements. Further, even if an AI machine could be named as an inventor, the patent applicant would still have to prove their entitlement to the patent and the Court of Appeal also held that there is no rule of law entitling the owner of an AI machine to intangibles produced by the machine.

The consultation notes that if AI-devised inventions are not able to be patented then there may be less investment in AI technology, or, it may encourage the use of trade secrets, which could harm follow-on innovation. However, if AI-devised inventions are permitted, this may lead to a proliferation of patents if AI reduces innovation costs, and these could potentially be held by a small number of dominant entities with the best AI technology and training data.

Policy proposals

The consultation presents four policy options:

•    Option 0 – no legal change. This option would maintain the position that patent protection is only available if the inventor is a human. Patent protection would remain available if AI is used as a tool by a human inventor. A patent would not be available if AI is named as the inventor or co-inventor of an invention.

•    Option 1 – the definition of 'inventor' expanded to include humans responsible for an AI system that devises inventions. Under this option a patent application would still have to name a human inventor. However, the law would clarify that an inventor for a AI-devised invention (that is, one where no human can be identified as the inventor) would be the human who made the arrangements necessary for the AI to devise the invention. The present test for the "inventor" – the actual deviser of the invention – would not be altered. Patent entitlement would flow as under the current law. Under this approach, people could be considered inventors if they were involved in: programming the AI, configuring the AI, operating the AI, selecting input data as training data, or recognising applications of the output of the AI.

•    Option 2 – allow patent applications to identify AI as an inventor. This option would necessitate the patent system being amended to allow UK patents to protect inventions devised solely by AI. There would be no requirement to name a human inventor. The proposal suggests two ways of achieving this: amending the legislation to allow AI to be named as the inventor, or  amending the legislation to remove the requirement to name an inventor if the invention is devised by AI.

Under this proposal, the human "closely responsible for an invention devised by AI" would be entitled to the patent in the first instance. If no human qualifies as inventor, the patent owner would be the human who made the arrangements necessary for the AI-machine to devise the invention. This approach would not confer on AI systems the right or ability to apply for or own patent rights. It would also not change the well-established 'inventor' test. However, it may require legislative amendment to allow for legal challenge if AI has been wrongly named (or not named) as the inventor.

•    Option 3 – protect AI-devised inventions through a new form of protection. Perhaps the most interesting option suggests creating a new form of protection for AI-devised inventions.  This would create a right similar to a patent but with more limited exclusive rights. Any such right would operate alongside the existing patent system. The consultation notes that the relationship between the two systems would need to be clearly articulated, especially where humans and AI are co-inventors.

This proposal suggests that the new right could have similar conditions for grant as for human-devised inventions but also that it may need a stricter test of inventive step as AI may invent in ways human inventors would not consider obvious. Alternatively, the right could be granted with no test for obviousness, with novel AI-devised inventions being automatically protected and decisions on validity left to the courts. In either case, it is suggested that protection could be less than the 20 years conferred by patents.

The consultation acknowledges that if a shorter duration of protection is offered, this might encourage applicants not to acknowledge AI inventorship, or the converse if the new right is quicker and easier to obtain. As this hypothesised type of protection does not appear to be offered in any other country, the consultation also seeks evidence to show why it would incentivise innovation and investment in AI.

Ranking and next steps

As with the part of the consultation concerning copyright, respondents are asked to rank the proposals and are also asked whether they might have any consequential effects on the existing patent system. With respect to Options 1 and 2, the consultation seeks opinions on how the inventor of AI-devised inventions should be identified and who should own the patent. It also canvasses views on whether respondents' global patent filing strategies would be impacted by the changes proposed in these options. Regarding Option 3, the consultation seeks views on the term and scope of protection such a new right would provide, and what criteria should be used to assess grant and why.

The consultation remains open until 7 January 2022 and responses can be made online. The responses will be used to inform government decisions on any changes to the existing law in this area. However, there is not yet a timeframe for when a policy decision can be expected.

This article is part of a series on the future regulation of AI and the UKIPO consultation, which also includes a look at the latest proposals related to copyright and text and data mining.