The U.S. Supreme Court's decision in Alice earlier this decade has closed the door on many drug and device patents involving artificial intelligence (AI), as such claims are likely to be dismissed as abstract ideas, which means they can't get through what's become the patent eligibility rabbit hole of Section 101 of the Patent Act.
"The current state of the law on 101 is a pretty big hurdle for evolving technology," Nimalka Wickramasekera, a partner at Winston & Shawn LLP, told BioWorld MedTech. The court's Alice test for 101 is even more challenging for AI, since people are still wrapping their heads around what they're claiming as the invention, she added.
"It's a particularly perilous time for those trying to prosecute their patents," Wickramasekera said. At the same time, AI is becoming increasingly important in medical devices, as it allows them to become more adaptive.
It also is becoming a valuable tool in diagnostics and drug development. For instance, AI can be used to select a molecule and then predict how well it will work in practice and how it will react with excipients and other drugs, said Stacie Ropka, a partner at Axinn, Veltrop & Harkrider LLP. And it can combine the diagnostic abilities of hundreds of specialists to work out what's ailing a specific patient.
Testifying before a Senate Judiciary subcommittee in June, Laurie Hill, vice president of intellectual property at Roche Holding AG's Genentech Inc., also touted the possibilities of AI in the drug and device sectors. "One important transformation on the horizon is the rise of bioinformatics, in which biotechnology and AI are brought together to inform all stages of personalized medicine, including medicine development, diagnostic development and patient treatment," she said.
Hill explained how Genentech is using AI in developing nucleic acid sequences to encode a portion of a patient's own tumor to create a personalized medicine to stimulate the immune system to fight the tumor. AI also is being used "to inform clinical trial design, leading to innovative trial designs and analyses that promise to reduce the cost of clinical trials and to expedite product approvals," she said.
The tricky part is getting a patent on all those innovations. "101 is, of course, problematic," Ropka said. While the Supreme Court has said abstract ideas can't be patented, it hasn't defined what constitutes an abstract idea, she told BioWorld MedTech. Consequently, the lower courts and the U.S. Patent and Trademark Office (USPTO) have applied Alice unevenly. The USPTO has tried to fix the problem, but there's still no clarity on what an abstract idea is.
Thus, Wickramasekera said she wouldn't be surprised if every patent application based on AI has had at least one rejection at the USPTO. To get around the court's prohibition on patents for abstract ideas, AI inventors must go beyond the algorithm and its application in computing devices, she added. The claims must be concrete but adaptable.
That may be easier said than done. "We have experienced regular rejections from the [USPTO] on Section 101 grounds, even when our inventions pass all other sections of the Patent Act and clearly merit protection," Hill told the Senate Judiciary Subcommittee on Intellectual Property. These are inventions that are patentable in other countries, she noted. (See BioWorld MedTech, June 13, 2019.)
The leaders of the subcommittee – Sens. Thom Tillis (R-N.C.) and Chris Coons (D-Del.) – are working on a bill to clear the confusion left in the wake of Alice and a trio of other Supreme Court decisions impacting patent eligibility.
When a patent is rejected, the applicant can try to amend its claims to overcome USPTO objections. But in writing around the 101 rabbit hole, the applicant risks narrowing the scope of the claims so much that they have little value. "It's like giving the farm away," Ropka said.
In a quick search of drug patent applications involving AI, Ropka found two, neither of which had been granted. One involved the use of AI to select a drug, be it a biologic or small molecule, based on a specific binding capability. The patent was denied and then appealed; the appeal has been on hold since 2017, she said.
The other one claimed a computerized method to characterize a molecule and then compare one well-characterized region of that molecule with a database of other known regions. The patent was rejected. The applicant tried several times to write around the 101 objections before finally abandoning the claims, Ropka said.
In urging Congress to fix the problem the courts have created, Hill reminded the subcommittee, "It is important to ensure a stable and certain environment for investment in the future of innovation, which is likely to include more and more innovation involving AI."
Whether Congress has the appetite to address the 101 patent wonderland is debatable. Lawmakers who characterize biopharma as the epitome of big corporate greed often point to the patent thickets companies amass to protect blockbuster drugs. Several of them have proposed limiting the number of patents on a drug or making it harder to get patents. They might be leery of passing a bill that could enable more patent protection.
Even if the bill Tillis and Coons are developing were to become law, there's no guarantee it would be sized to fix the problem. It could simply codify the patent exemptions the Supreme Court has shaped. If it does that without defining what can't be patented, it "won't do anybody any good," Ropka said, because there still will be uncertainty about what's patentable. Biopharma and med tech need to know the risks and strengths of 101 issues, she added.
Challenges for the future
While 101 is the threshold question for AI, more questions are starting to surface as AI emerges from the other side of the looking glass. Perhaps at the top of the list is what happens when AI produces new inventions. "That's absolutely going to be the question we're grappling with," Wickramasekera said. "Who's the inventor?"
Patent law only recognizes people as inventors, Ropka said. If AI is considered a tool, would the inventor of something developed through AI be the person who came up with the algorithm? Or the person who selected the data to feed it?
And what about an invention that changes over time? Noting a guidance the FDA is developing on medical devices that will adapt through AI, Wickramasekera said the agency recognizes that the device it approves won't be the device that's marketed in the future. In such a case, the question of who's the inventor of the future iteration of the device could lead to litigation.
Patent attorneys are already pondering such questions, and Wickramasekera expects to see some litigation around these issues in the health care sector in the next two to three years and possibly sooner in the high-tech field.
Another patent challenge AI could present in the near future is obviousness. As it becomes more prevalent, AI will be the prior art, Ropka said, as it is creating the proverbial "hypothetical person skilled in the art who knows everything."
When that happens, getting a patent for new inventions could become even more challenging.
What is Alice?
In its unanimous 2014 decision in Alice Corp. v. CLS Bank International, the Supreme Court made it more difficult to patent inventions stemming from "abstract ideas," such as algorithms. Abstract ideas, by themselves, cannot be patented; they must have an inventive concept underlying them. Merely requiring computer implementation of an abstract idea fails to transform it into a patent-eligible invention, the court ruled.