Most businesses have filed, or are planning to file, an artificial intelligence-related patent application but few are certain of how to, or whether they can, get patent protection for these products, according to a survey of more than 60 in-house patent lawyers.
Patent Strategy surveyed corporate lawyers from different sectors, including life sciences, technology, chemical, telecoms, manufacturing and automotive, and various jurisdictions to reveal the business concerns around patenting AI-related products.
The data showed that just under half of in-house counsel (44%) had filed a patent application for an AI or AI-enabled product, and another quarter were planning on filing one within the next five years; only 11% said they would never file one.
When does your company intend to file a patent application for an AI OR product making use of AI?
But when asked what their biggest challenge to getting patent protection for an AI product was, 57% of respondents cited ‘excluded subject matter’ (31%) or a lack of internal or external counsel expertise in the field (26%).
Fewer in-house lawyers were as concerned with the more arduous challenges of finding prior art in non-traditional places (13%); wading through the vast number of AI filings to determine freedom to operate (8%); or having to wait during long grant processes (3%) – although counsel say these are still important issues.
What is (or will be) the biggest challenge when seeking to patent an AI or product making use of AI?
Sources say that this data is reflective of the lack of knowledge between businesses, law firms and patent offices, when it comes to patent rules for technology; despite its growing prominence. They add that businesses – in particular traditionally non-tech firms – often do not have sufficient technical knowledge of the technology, and that patent offices have not sufficiently interpreted and clarified software patentability laws.
“There is a mutual lack of knowledge between patent offices and businesses,” says the head of IP strategy at a manufacturing company. “Patent offices are not sure where to draw the line on what is patentable and what is not and many businesses do not have the necessary technical knowledge of the technology.” He adds that his firm has had success in patenting AI products but only because it is lucky enough to employ a patent attorney who spent several years at a large software company that is a major AI patent filer.
A senior patent lawyer at a sporting goods company says that one reason why businesses may not have the right knowledge is that AI has only recently emerged as a viable technology. He points out that most senior in-house patent attorneys would not have had a chance to study AI at university because there was not the same focus on that technology as there is now.
Carlos Rosario, IP attorney at Peloton Technology, agrees that there is a mutual lack of understanding, and adds that excluded subject matter is a common problem when it comes to patenting AI-related products at the USPTO.
“A lot of software has been shut down in the US lately, although that may change. The US Patent and Trademark Office just came out with due guidance last month,” he says.
“I draft patents and describe different types of AI systems but I am not an expert in the field and if I was I would probably be an engineer or external counsel. AI is not easy to learn, depending on the type – not so much the application but when it comes to describing how a system works. That can be very difficult.”
Patent offices have traditionally been resistant to granting software-related patents because of exclusion law. Article 52 of the European Patent Convention (EPC), for example, says that inventions shall be granted in “all fields of technology” but that some subject matter is not eligible for protection when claimed as such. That matter includes scientific theories and mathematical methods, aesthetic creations and programs for computers – which is a problem for AI patentability because the technology is seen by some as pure mathematical methods and nothing else.
Guidance needed
Survey respondents were clear about what they expected from patent offices to alleviate the uncertainty surrounding AI patenting. Almost two thirds of in-house lawyers (63%) said that the most important improvement these organisations could make to ease the patenting process was to issue guidance on AI patenting laws, making use of pointers and examples.
More than a quarter of respondents (27%), however, said the best improvement that patent offices could make to ease the AI patenting process was to work with them to better understand their filing challenges. Few respondents (5%) thought that patent offices expanding their AI–knowledge base was the way to go and no one suggested a different improvement to the three listed in the survey.
What is the most important improvement patent offices can make to ease the AI patenting process for businesses?
“In India objections are constantly raised under Section 3k of the Indian Patent Act, which means it is hard to file software patents as such. But if you draft the claims in a particular way or manner, you can file such a registration,” says the IP strategy manager at an automotive company.
He adds that if the drafter is well aware of case studies of successful AI patent applications that include details on how and why it was granted, gaining patent protection should not be a problem.
“Interpretation of the laws and the impact that they have on the way that attorneys draft patents is very important. If businesses knew how to draft claims that illustrate technical steps, by referring to detailed guidance, patenting software-related products would be easier.”
Rosario at Peloton adds that the Section 101 debacle over patent exclusion for business methods, such as stock-predicting algorithms, has been driving patent practitioners and examiners crazy for four years, and the situation would benefit from good guidance containing relevant examples and pointers.
“It would be very helpful to know what is patentable, and what is not, when it comes to AI,” he adds.
The sporting goods senior patent lawyer says patent offices have started to engage with the topic of AI but many are reluctant to publish guidance because little case law has been generated on the matter compared with other areas, and few want to risk publishing inaccurate pointers or tips.
The desire to see patent offices issue AI guidance is likely to be why 71% of survey respondents said that the European Patent Office (EPO) offered the best service for patenting an AI or AI-enabled product, out of the big five patent offices, based on factors such as examiner leniency in practice, application ease and guidance.
The office published guidance that came into force in November 2018, where it outlined that an inventor must show the claimed subject matter has a technical character (although it may contain non-technical features) and that all features contributing to the technical character are taken into account for assessment of the inventive step.
The results may have been influenced by the fact that most respondents to this survey were based in Europe. But it is worth noting that around half of North American respondents cited the EPO for best service when it came to patenting AI, while approximately 15% of Europeans said the same for the USPTO.
“The fact that so many businesses want to see guidance from patent offices is not surprising and it is why the EPO has done so well in this question,” says the head of IP at an energy company. “The EPO has a consistency of applying the rules to a problem that is better than the other offices.”
A US patent attorney at a global pharmaceutical manufacturer adds that the EPO has taken the lead among patent offices in making the AI examination landscape more inviting.
“It is the political issue of AI being the hot area of computer science right now and making the patent landscape around it friendlier to spur economic growth. The Europeans recognised that before the US and the PTO is now trying to catch up.” He adds that he attended a public policy meeting in Washington DC recently where there was a recognition that improvements were needed to ease the technology’s patenting process.
The head of IP strategy at a manufacturing company agrees that the EPO has the best potential for making AI patenting easier. He adds that the USPTO has a strong history for patenting digital innovations, which is likely why 24% of respondents said it offered the best service. The PTO’s traditional status in the digital world, he says, is why the EPO has tried to make themselves more attractive to businesses looking to commercialise AI.
Which patent office currently offers the best service for patenting an AI or products making use of AI (based on factors such as examiner leniency in practice, application ease and guidance)?
It is worth noting that the Japanese Patent Office (JPO) has issued similar guidance to that published by the EPO, but was not cited by a single in-house lawyer for the best service. This result may have occurred because no respondents were based in Japan, and consequently votes that would have gone to the JPO for its efforts were snatched up by the EPO.
Machine yearning
Considering the lingering uncertainty in the AI patenting space, businesses may choose to rely more heavily on their external counsel for advice – particularly the quarter of our survey respondents who will file for an AI patent in the next five years.
An IP attorney at a global automotive company, which is a key autonomous vehicle filer, points out that while businesses will seek to have the necessary skillsets in house, patent departments have limited resources and will likely choose to outsource a lot of work on technologies that they’re less knowledgeable about.
“We use external counsel for quite a few of these AI and tech-related matters to help move them along,” he says.
With that in mind, law firms should have an understanding of what clients and potential clients want to see from external lawyers when it comes to patenting AI-related products.
When given the opportunity to say what external counsel could do regarding patenting an AI, or product making use of AI, 37% of respondents chose to write down their thoughts; most had the same sort of comments.
Most in-house counsel said they wanted private practice firms to improve their technical knowledge of the subject to draft claims better and advise businesses on whether an application for a particular AI product would pass examiner scrutiny.
“External attorneys are more familiar with the legal framework, but less familiar when it comes to products to be covered by the claims,” wrote the patent counsel at an automotive firm. “They should inform themselves about the market and the industry in general.”
The head of IP at a European university also wrote: “[Law firms] should advise on when patent protection is appropriate or when it is, in fact, more of a 'software only' or know-how invention. In other words, they should know when to not file a patent.”
Responding to these comments, the pharmaceutical firms’ US patent attorneys adds that it is important for external counsel to act as the bridge between what the normal in-house attorney thinks is patentable and the realities of the machine learning space.
“Outside counsel can, or should be able to, speak the language of data scientists more readily than us. They should be there to help us recognise and assess the invention.”
Patent Strategy’s landmark survey shows that despite the growing prominence of artificial intelligence across different industries, most businesses have yet to attain real legal certainty when it comes to protecting their lucrative inventions. Patent offices are making impressive strides in the area – particularly the EPO – but clearly more needs to be done.
Law firms can help in this process but need to make sure they’re properly advising their clients – particularly on whether or not patent protection is even worth pursuing on a particular AI product.
Part two of Patent Strategy’s special report on AI, which reveals what IP departments think about patenting inventions discovered by an artificial intelligence, will be published next week. Subscribe to the newsletter to receive it.
