By: Joe T. Schuler, Esq. and Teige P. Sheehan, Ph.D.
The Court of Appeals for the Federal Circuit recently handed down what is likely to be the first among many decisions shaping the patentability issues surrounding artificial intelligence. In Recentive Analytics v. Fox, the Federal Circuit leaned on existing law applicable to eligibility of software for patenting, reasoning that the patent at issue, as the District Court put it, claimed the use of “known generic mathematical techniques.” In the case, Recentive brought infringement claims against Fox based on Fox’s use of AI to create network maps and schedules. The claims in Recentive’s patents were not specifically to the machine learning itself, but only to the use of AI to create the network maps and schedules. Indeed, the claim language itself was to “any suitable machine learning technique.”
The Federal Circuit determined that the use of AI to create maps and schedules, even ones that adjust to updated “real-time” data, is essentially the same as using the computer “merely as a tool,” and thus claimed an unpatentable abstract idea performable by humans. The court also determined there was nothing novel about iteratively training the AI. After all, that is already known as one of the core features of AI.
While the court held the claims to using AI in this way are patent ineligible, its characterization of the patents as simply claiming AI use as a tool was important to the decision. The claims at issue were to the use of AI (a version of software) to perform tasks that historically network employees had to do either by hand or with less sophisticated computers. To those familiar with the court’s attitude towards such subject matter in recent years, from a high level the holding was somewhat predictable. The Federal Circuit did not rule that claiming use of AI is per se ineligible, meaning other patents with such claims may be found eligible. Perhaps the Court recognized this, acknowledging in conclusion that, “[m]achine learning is a burgeoning and increasingly important field and may lead to patent-eligible improvements in technology.” Notably, a petition for rehearing by the Federal Circuit, filed by Recentive, is currently being considered.
A different but related question is whether something generated through the use of AI is patentable. In Recentive, the patents at issue claimed the use of AI. But in other cases, inventors may use AI to assist in creating claimed inventions. Whether inventions created with the use of AI are themselves patentable also remains an open question, separately from whether and when use per se of AI is patentable. For example, AI is already being used to identify drug targets in the pharmaceutical space, predict bespoke Cas9 proteins suited for therapies, assist in drug discovery generally, and predict molecular binding affinity. These capabilities allow researchers in the pharmaceutical space to invent drugs with a greater rate of success, but it introduces as-yet unresolved legal questions surrounding inventorship and patent eligibility. Some of the software models used to assist in generating claimed inventions are open source, allowing many companies to use them at low cost. The end result in many of these circumstances satisfies some of the requirements for inventiveness (i.e., novelty, non-obviousness), but the question is whether, how, and under what circumstances the fact that AI created or at least assisted in the inventive process should impact patentability, if at all.
Patent law is responding to the emergence of AI and its impact on technology, both with regard to protecting inventive uses of AI and inventions created by the use of AI. Case law, legislation, and executive branch policy will likely shape the direction of both. The U.S. has been leading the way globally in the development of AI; the lion’s share of patent applications related to AI have been filed within the last 10 years.[1] It will take time for courts, policymakers, and inventors to decide how to handle protecting AI and resolve the disputes the issues generate. Developers of new AI applications and tools would be wise to attend to the developments in the law as they arise so they can best protect their intellectual property.
[1] https://news.bloomberglaw.com/us-law-week/patent-litigation-trends-in-artificial-intelligence?context=search&index=3