BIO 2020: AI patents and secrets
Discussions around artificial intelligence (AI) and IP can sometimes be futuristic and trapped at the level of high-concept: whether machines can be inventors, for example. But at BIO Digital 2020, a panel of experts was on hand to provide a much more practical, down-to-earth guide to patenting AI inventions.
AI, after all, is becoming an increasingly everyday technology. This is usually in the form of machine learning (ML)—the “hot and popular way of doing AI”, as Greg Silberman, deputy general counsel at BlackBerry, described it.
While AI and ML source code can be protected by copyright, and be the subject of trademark registrations, most of the discussion on AI and IP usually deals with patents. The central question of the panel discussion was, when it comes to AI, “to patent or not to patent?”.
Before answering that question, the panel looked firstly at “what can you patent?”. This has by no means been a simple question in the context of the US over the past decade. As Carl Kukkonen, partner at Jones Day explained, there are numerous everyday applications of AI and ML eligible for patent protection, especially in the life sciences.
These include drug discovery tools that can help identify the right targets, medical image processing (used in fields like radiology, surgical imaging, and dermatology), and personalised medicine, among other products.
But the possibilities are not endless. As David Kerwick, legal director at Merck, Sharp & Dohme explained, there have been numerous “limitations on what you can patent” in the US in recent years, of which the US Supreme Court’s 2012 Mayo decision is just one.
You can’t, notoriously, patent laws of nature, abstract ideas, or natural phenomenon, and proving your invention turns these “building blocks” into something more than that can be difficult.
As well as that, there’s the risk of infringing others’ IP. That’s what makes freedom-to-operate searches so important, especially in the life sciences, explained Donna Robertson-Chow, senior director at Dorf & Nelson.
“We often get asked to carry out freedom-to-operate searches in the life sciences field because life sciences companies put so much effort, money, R&D, resources in early on,” she said.
“You don’t want to spend ten years in a R&D project only to find out that there was an earlier patent out there,” added Robertson-Chow.
These are all key considerations to bear in mind when devising a strategy to patent your AI invention. But is patenting always the best choice? Returning to the question of “to patent or not the patent”, the panel considered whether trade secrets could be just as valuable a pathway for AI companies.
In Silberman’s opinion, certain back-end “know-how” should be withheld as trade secrets. But with trade secrets comes the risk of misappropriation.
As Kukkonen pointed out: “trade secret identification is becoming increasingly important. So if you want to assert a trade secret misappropriation case, you need to identify your trade secrets in the complaint. It’s helpful to have an internal trade secret identification process to help protect you in future.”
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox.
