Patenting AI: the IP office perspective


Alex Baldwin

Patenting AI: the IP office perspective

peshkova /

As the number of AI patent applications grow worldwide, IP offices have the difficult task of granting or refusing protection.

The US Patent and Trademark Office (USPTO) estimates that the number of AI patents it has received per year has doubled in two decades, reaching 60,000 in 2018.

With the encroaching ‘fourth industrial revolution’ bringing with it advancements across every facet of the tech world, it is the lawmakers that need to keep pace to facilitate the expansion.

In a special session of WIPR Patents Live made possible by Morningside, three AI patent specialists discussed the patent examiner’s perspective, a recent European case, and other important legislation in the Patentability of AI-Driven Devices panel yesterday.

Kicking off the panel, John Harris, Morningside’s European director of business development, IP, said the global AI market is “forecast to hit $327 billion this year and $500 billion by 2024”.

Challenges for examiners

However, patenting AI inventions is difficult. “Patent eligibility is a known challenge,” said Rob Bahr, deputy commissioner for patents at the USPTO.

“AI inventions are treated just like any other computer-implemented invention (CII)… They may raise issues if they fall into the ‘abstract idea’ exception but these can be avoided in the drafting stage.

“If you recite steps that cannot just be performed mentally, integrate abstract ideas into a practical application and then recite features that relate to an identified improvement, then your idea is likely not abstract.”

Bahr added: “Patent disclosure is also an issue, there aren’t a lot of cases around disclosure requirements for AI.”

Across the Atlantic, Abderrahim Moumen, the European Patent Office’s director of operations, information and communications technology sector, gave an examiner’s perspective in tackling complex AI applications.

“Applications sometimes have very complex mathematical claims and require examiners with particular expertise, so you need to ensure people are trained and making sure you have the right mix of competencies,” said Moumen.

“We are also seeing an increased amount of prior art, mainly from Asian documentation or non-patent literature which need to be considered, looking at how best to access this art is a challenge.”

Reactions to G 1/19

One of the biggest rulings of this year laid out, in Europe at least, clearer guidelines on AI patentability.

The G 1/19 decision last month, handed down by the EPO’s Enlarged Board of Appeal, held that the established case law on CIIs also applied to computer-implemented simulations.

“In general we are pleased with the outcome,” said Moumen. “When we started, we wanted a distinction between software development and technical innovations. It is about what happens when the software is run, a technical aspect needs to be considered for patentability—G 1/19 has confirmed that this is the right approach.”

“G 1/19 was incredibly important for IBM,” said Anita Shaw, dual-qualified European patent attorney and chartered UK patent attorney at IBM.

“We were tracking the case and even sent an amicus brief related to the proceedings. We are also pleased with the outcome confirming the status quo.”

Changing laws

From an inventor's perspective, G 1/19 spells greater clarification around AI invention patentability, but do more steps need to be taken to ensure a better understanding? Patents Live asked the panel if there is a need for legislative solutions to treat AI separately from other CIIs.

“The current patent systems are fairly robust,” said Shaw, “We would encourage examiners to look at these new inventions through existing policies before introducing anything new. Let's go slow before we start legislating around this.”

Bahr agreed: “I would add that the USPTO  has had a lot of stakeholder engagement around legislative changes. They said the current legal framework was satisfactory for handling these patents.”

Alternative protections

While the case law and prior art around AI continues to grow, companies might want to rely on other forms of protection in some scenarios, such as trade secrets.

“The discussion around sufficiency and disclosure of AI patents speaks to the multiple routes you can take to protect the IP,” said Shaw.

“Patents are pro innovation and help fuel inventors. They are a dissemination of technical innovation. Trade secrets are the opposite of that. There may be scenarios where that is the right protection over patents.”

Watch ‘Patentability of AI-Driven Devices’.

WIPR Patents Live hosts weekly broadcasts from some of the best speakers in the technology sector in the form of virtual panel discussions, roundtables, webinars and presentations.

Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox

AI, patents, USPTO, Morningside, EPO, IBM