Inventive AI: UK Supreme Court holds that only humans can be inventors

The Artificial Inventor Project

Over the past few years patent offices and courts around the world have been grappling with the question of whether a patent can be granted when the inventor is an AI system rather than a human. This has been prompted by Dr. Stephen Thaler’s Artificial Inventor Project. Dr. Thaler has filed patent applications in numerous jurisdictions designating his AI system (known as ‘DABUS’) as the inventor. The UK patent office refused Dr. Thaler’s applications on the basis that an AI system cannot be specified as the inventor. This issue recently reached the UK’s most senior court, the Supreme Court, which concluded that the patent office was correct – under the UK patent’s regime inventors must be human.

Can AI be an ‘inventor’ under the UK Patents Act?

Dr. Thaler is seeking to generate debate around these issues. However, the Supreme Court made clear that the appeal did not concern whether inventions autonomously created by machines should be patentable, but rather whether the current provisions of the UK Patents Act allow for the inventor to be a machine. The Supreme Court therefore carried out a detailed review of the relevant legislative provisions. Both the specific provisions regarding entitlement to a patent and the provisions addressing ancillary issues supported the conclusion that the term ‘inventor’ in the Patents Act did not extend to a machine such as DABUS.

Was Dr. Thaler entitled to a patent for the invention made by DABUS?

In the UK, a patent may be granted to the inventor or to someone who derives ownership from the inventor (e.g. the inventor’s employer or assignee). While UK copyright legislation stipulates that the author of computer-generated works is the person who made the arrangements for the creation of the work, there is no equivalent in patent law.

Dr. Thaler argued that under the ‘doctrine of accession’ he was entitled to patent inventions made by DABUS because he owned DABUS. The ‘doctrine of accession’ provides that a new item produced by an existing item (such as a calf produced by a farmer’s cow) is owned by the owner of the existing item. The Supreme Court disagreed that Dr. Thaler would own innovations created by DABUS:

1. Any technical advances made by DABUS were not an ‘invention’ for the purposes of the Patents Act because an ‘invention’ within the meaning of the legislation must be devised by a person.

2. In any event, the doctrine of accession does not apply to inventions.

Validity of patents for AI-generated inventions - a less dogmatic position

The Artificial Inventor Project is a campaign to establish that AI systems can make inventions and that their owners are entitled to patents for such inventions. Dr. Thaler at all points insisted that DABUS had made the relevant technical advances autonomously. The Supreme Court emphasized that the outcome of the proceedings may well have been different had Dr. Thaler taken the position that he was the inventor and devised the invention using DABUS.

Whilst it may be straightforward for a less dogmatic applicant to avoid the pitfalls suffered by Dr. Thaler in the application process, the Supreme Court’s decision may leave open arguments relating to validity at a later stage. Where an application identifies the inventor as an individual that used AI as a tool, the patent office may not object, but if the innovation was actually devised by AI, a third party may be able to allege invalidity on the basis that an AI-generated advancement is not an ‘invention’ for the purposes of UK patent law. The Supreme Court confirmed that ‘[i]f patents are to be granted in respect of inventions made by machines, the 1977 [UK Patents] Act will have to be amended.’

Patent applicants should be mindful that inventions created by AI may face a greater risk of refusal. However, in many instances the research will be sufficiently human-led that a natural person can be specified as the inventor (notwithstanding AI’s contributions to the process). Dr. Thaler insists that DABUS should be the named inventor rather than himself, but others should be prepared to be more flexible in order to obtain protection.

Contracts for the use of AI in R&D

As AI distances the inventive process from the humans with responsibility for the research, identifying the (human) inventor and therefore entitlement to a resulting patent may be less clear-cut and may be more likely to give rise to competing claims from parties involved in the research. Contracts for the use of AI for R&D should pre-empt this uncertainty and include more detailed IP ownership provisions than might be expected in a traditional software licence or SaaS contract.

In addition, as AI becomes more responsible for inventions, owners of AI systems that make their technology available to support third parties’ research are more inclined to seek to negotiate ownership of inventions created by their AI – pitching themselves as a collaboration partner rather than a traditional IT supplier. An AI provider may propose alternatives to the subscription-based fee models traditionally offered by software providers (e.g. royalties in respect of the inventions that their AI systems create).

Comment

The UK Supreme Court’s conclusion is in line with the European Patent Office¹, the U.S. Patent and Trademark Office and the patent offices of numerous other jurisdictions². The UK government has been reluctant to update patent law to address machine inventions. Following a consultation covering the patentability of AI inventions, in June 2022 the UK government expressed the view that AI was not sufficiently advanced to invent autonomously and that the issue should be addressed in a harmonized way internationally. However, since then there has, of course, been a boom in the use of AI. In the U.S. new guidance is expected in the coming weeks as President Biden’s Executive Order on AI³ set a deadline of 27 February 2024 for the USPTO to publish guidance on the issue of AI and patentability. As AI becomes more sophisticated and human involvement reduces, legislative reform will be needed in the UK for AI-generated inventions to be appropriately protected.

The European Commission’s proposed Regulation on Artificial Intelligence is a major step in regulating the use of AI⁴. However, that proposed regulation does not address how the benefits of AI’s ingenuity, in particular IP, should be attributed. This will remain a matter for contractual negotiation. Where AI is expected to be central to the inventive process, the contractual terms need to account for the more prominent role of IT in the process.