Long read

Patentable or not? Regulating AI inventions for a digital future

The recent G 1/19 hearing at the Enlarged Board of Appeal has reignited the patent community's interest in the law on AI patents. Although concerning computer simulations, the scope of G 1/19 is such that its conclusion could fundamentally alter the patenting of AI-related or AI-driven inventions. For many patent attorneys, the development of guidelines on AI patents would be welcome. But the approach of patents courts and offices requires a careful balance of interest.

24 July 2020 by Amy Sandys

AI patents Recently the Enlarged Board of Appeal at the EPO heard a referral in the G 1/19 case concerning computer simulations. Its similarities to the tech behind AI means the future of AI-owned patents, and patents relating to AI functions, is under scrutiny ©_nastassia/ADOBE STOCK

Last week, the European Patent Office streamed its inaugural Enlarged Board of Appeal hearing via live video software. Around 1,600 people tuned in to the referral, which looked at the technical aspects of a patent application covering computer simulations of pedestrian environment. It’s almost 20 years since the owner first filed the patent and still there is no definitive answer. But the case’s outcome is intrinsic to the development of AI patents.

Abstract maths

The hearing sparked interest among a global patent community for whom patenting in the computer inventions and AI is gathering importance. Artificial intelligence belongs to the framework of computer-simulated inventions; both are based on abstract mathematical theory. As a result, so far there are no specific rules on patenting AI. Current case law in computer simulations is also applicable to AI patenting cases.

For some patent attorneys, existing law is sufficient. AI sits at the intersection between technical and non-technical patenting. They take the view that as technology continues to develop, so too will the legal and political response to AI patents. Furthermore, the technicality of an AI invention remains a fundamental part of its patentability.

Andreas Holzwarth-Rochford

Andreas Holzwarth-Rochford

Andreas Holzwarth-Rochford, partner and patent attorney at Jones Day, says “The EPO is willing to change and adapt. Currently, protecting AI is in a good situation because the office has the opportunity to fine-tune the existing case law, which will benefit some AI developments.”

On the other hand, some patent attorneys argue the current framework is insufficient to facilitate future innovation. They argue that change is needed at all levels to secure a future for AI patents.

In February 2020, the new European Commission president Ursula von der Leyen laid out her plans to regulate AI in the first 100 days of her presidency. The coronavirus crisis among other factors has halted progress. But at a national level, the conversation around AI continues to gather traction.

President to precedent

G 1/19 concerned a referral made by the EPO’s Technical Board of Appeal. The referral asked the Enlarged Board to clarify the case law on computer simulations under the European Patent Convention (EPC). This is due to a discrepancy over a similar case regarding Infineon Technologies (T 1227.05).

The original application, EP 15 46 948, is entitled “Simulation of the movement of an autonomous entity through an environment”. The Enlarged Board focused largely on whether computer-simulated inventions need a ‘direct link with physical reality’ or if having ‘technical effect [and] purpose’ is sufficient. This meant the Enlarged Board considered what is patentable in the realm of computer simulation technology, and to what extent the technical aspects of an invention can be considered.

Pawel Piotrowicz, AI patents

Pawel Piotrowicz

Pawel Piotrowicz, the patent attorney behind G 1/19, argued that the Boards of Appeal should uphold its approach in Infineon to patenting software inventions.

In upholding this approach, Piotrowicz argued, the EPO would enable future innovation in the software patenting and AI fields.

Impacting the every day

But this approach would not just benefit communication technology, says Piotrowicz. Adopting a less restrictive approach would allow other industries, such as pharmaceuticals, to use computer simulations in developing new applications.

Piotrowicz says, “The outcome of the G 1/19 case will have an impact on the assessment of patentability of not only simulations, but also potentially other forms of computer-implemented inventions, such as machine learning and artificial intelligence.”

“The decision will affect whether software tools used in designing everything from bikes, to airplanes and computer chips, to the internet are patentable,” he says.

Holzwarth-Rochford says, “Now in all technical areas applications and uses for AI exist. Mechanical engineering, for example, uses AI in developing the form of a ship or a car. It’s also used in material sciences, computer sciences, biology, pharma and chemistry. Lots of research and development organisations as well as governmental initiatives like ‘Industrie 4.0’ are pushing companies into exploring these areas.”

So while the G 1/19 decision is not concerned specifically with AI, it carries weight for the future of AI patenting. Both computer simulations and AI are based on computational models and algorithms. The similarities between the two mean that much existing case law is applicable to both fields. But here is where the specific application of the AI system because important.

Economic imperative

Matt Hervey, partner and head of AI at the London office of Gowling WLG, says, “The law and procedure regarding inventions by AI seems clear. At least for the EPO, the UKIPO and the USPTO, these did not envisage AIs as inventors and require a human to be named as inventor.”

Matt Hervey

Matt Hervey

“But, long term, will a lack of protection for inventions by AI fail to reward research and development and lead to more inventions kept as trade secrets? Should patents be extended to inventions by AI? And/or should new forms of monopoly be created for such inventions?”

In some sectors, such as life sciences, a supplementary protection certificate (SPC) is an alternate means of asserting inventor rights. SPC regulation can extend the life of a patent for up to five years, because it can take a while for many pharma products to obtain marketing authorisations.

In turn, the delay can significantly shorten a patent’s usual 20-year lifespan. An SPC therefore straddles the middle-ground between patent law and regulation; some patent attorneys argue that future similar provisions must be made for AI patents, or AI-related patents. This is so inventors can reap the rewards of time invested in developing this new horizon of technology.

Alternative approaches to regulating AI is especially pertinent in areas such as pharmaceuticals, which increasingly apply to real-world development such as disease modelling. But often it is the inventor who must decide whether regulation is the correct route to take.

New routes

For example, Virginia Driver, partner and patent attorney at Page White and Farrer, explains that sometimes the value of an asset is such that asserting a monopoly is not necessarily the answer. “Inventors of AI, or AI-related patents, need control over what to do with their ideas in order for them to benefit humans and the economy,” says Driver.

Virginia Driver

Virginia Driver

“Of course, AI needs protecting. But patent holders should be willing to give open licences, albeit with restrictions around who can obtain them. This gives society the best chance of using AI for good.”

Conversely, Holzwarth-Rochford goes back to the notion of trade secrets, suggesting they are in fact a reasonable approach to protecting AI.

Especially since the introduction of the EU’s trade secrets directive, Holzwarth-Rochford says, “Trade secrets are an important tool to protect aspects of new developments especially in the AI sector, like algorithms.”

“This is especially because trade secrets can be applied in non-technical matters.”

He continues, “Unlike patents, trade secrets hold no restrictive timeline. It might be worth inventors considering this as an alternative or supplementary to the patenting route for the respective aspects of a new development.”

Waiting for Dabus

G 1/19 is not the only well-known patent case with the potential to impact future AI patenting case law. In January 2020, the EPO rejected two AI patent applications which designate the inventor as a non-human entity known as Dabus. The EPO rejected the applications on the grounds that AI, rather than a human, is listed as the inventor. In fact, the United States Patent and Trademark Office (USPTO) and UK Intellectual Property Office (UKIPO) have also both rejected patent applications for Dabus.

But last week, the UK High Court heard a referral from the UKIPO due to the potential importance of Dabus on future patent law. It is likely the UK parliament will have the final say on whether the IPO can award the application. At government level, such developments are crucial in setting a precedent for future inventions. Indeed, the IPO referred the case because deciding on the patentability of Dabus could fundamentally impact the future law on AI patents.

Thus, the future of AI patents is seeing movement. Hervey says, “Internationally, similar approaches are taken to core exclusions to patentability, for example in computer programs and mathematical methods. So the position of inventions covering AI is tolerably stable.”

At a global level, over July, patent offices such as the EPO and WIPO took part in consultations regarding potential future legal developments. The offices have discussed what future IP law should look like. This is not only for patent, but also for copyright and data protection. All aspects of IP are impacted by AI developments.

Furthermore, patent offices are beginning to utilise AI tools internally. An example is when searching databases for prior art. The application of AI extends beyond industry level, and into the matrix of patenting.

Model matrix

One needs only to read the industry’s Amicus Briefs submitted for G 1/19 to see an almost unamious push for AI patentability. However, Holzwarth-Rochford argues that, for the EPO, it is not so straightforward. “I would consider the EPO as generally customer-friendly,” he says. “But it also comes down to a balancing of interests.”

“Applicants want to ensure they can protect their technology, but the EPO must comply with the existing regulations and see if and how this is possible.” As Driver also says, “Any decision towards the patenting of AI without defining the practical context by the EPO would be a jump-step in patent law development.”

On the other hand, some patent attorneys view the European patent system as too rigid. For those on this side of the fence, the current framework means some emerging forms of technology are not protected. Thus, AI patentability is unfit for a digital future. Piotrowicz says, “If the EPO Enlarged Board of Appeal were to confirm that the legal framework is flexible enough to provide protection in these new areas, many would see this as a positive result.”

However, a development in regulations on AI patentability is not just down to various patent offices. Hervey suggests that the question is not, is the EPO ready? Rather, it should be at what level change is required. “Any reform of patent law to cater for inventions by AI, or to expand the scope for patenting AI, would require legislative changes,” says Hervey.

“That is likely to involve changes to national laws and international treaties based on economic analysis of promoting investment in inventive AI.”