AI and the Word that’s Been Missing from the Patent Eligibility Case Law
“Alice Step 2 at the USPTO has recently become a vestigial remnant that poses great danger for modern AI inventions.”
The artificial intelligence (AI) revolution poses new problems for deciding patent eligibility, problems for which the current body of U.S. Court of Appeals for the Federal Circuit case law and U.S. Patent and Trademark Office (USPTO) policy is ill-equipped to address. In particular, Alice Step 2, one of the most misunderstood doctrines in all of patent law, has the potential to become even more muddled when considering AI inventions. This is because the case law, as well as examiner practice, have tended to over-emphasize the importance of the conventionality or genericness of computers recited in the claims or described in the specification.
This state of affairs poses two converse problems for addressing the eligibility of AI inventions. First, groundbreaking AI inventions can be executed on generic computers. Thus, such inventions, which should be eligible for patent protection, should be entitled to claims that encompass completely conventional and generic computers. Conversely, inventions that are nothing more than abstract ideas can be, and often are, claimed as being performed by highly specialized machine learning accelerators in an attempt to perform an end run around Alice Step 2. In other words, such ineligible inventions should not be entitled to patent protection simply for reciting specialized processors.
In both cases, the conventionality or genericness of the computing hardware is largely beside the point. Instead, in order to adequately protect the most groundbreaking inventions of our time, new tools and guidance are needed so that the “search for an inventive concept” at Alice Step 2 does not devolve into a perfunctory “search for a generic computer.”
The Alice Steps
As background, deciding patent eligibility according to the Supreme Court’s 2014 Alice Corp. v. CLS Bank International decision involves a two-step inquiry. Under Alice Step 1, a claimed invention is evaluated to determine whether it is directed to an ineligible concept, such as an abstract idea. Abstract ideas include things like mental processes, business and commercial activities, and methods for organizing human behavior.
If a claim is directed to an abstract idea under Alice Step 1, there is a second chance to demonstrate eligibility under Alice Step 2. The claimed invention is then re-evaluated through a “search for an inventive concept” to determine whether there exist claim elements that are sufficient to ensure that the claim is not merely a “drafting effort” designed to monopolize the abstract idea itself.
This “drafting effort” language from Alice is a recognition of common patent prosecution strategies. Without trying to sound self-congratulatory, competent patent attorneys are clever—they have extensive experience dressing up mundane claims to seem highly technical. For example, rather than having a claim say “sending funds from a first party to a second party,” which sounds like purely ineligible commercial activity, the same concept can be claimed by reciting steps of retrieving values from a database, sending representative data over a network to a server system, parsing the data by the server system, transforming the data into a second format, and storing the transformed data in a user database.
A primary takeaway from Alice is that these two strategies for claiming the same concept should not be treated differently when assessing eligibility. And Alice Step 2 is supposed to be the vehicle by which a claim is judged to determine whether there are inventive features that are deserving of patent protection or whether the claim is merely a drafting effort designed to monopolize an abstract idea.
A Strange—and Dangerous—Evolution
However, in recent years, something strange has happened to Alice Step 2 at the USPTO when applied to inventions that involve computers, including all AI inventions. Namely, the inquiry at Alice Step 2 often merely asks whether the claims recite using a generic or a conventional computer. If so, the claims are ineligible.
This is a strange test indeed for several reasons. First, testing for the existence of a generic computer is vastly over-inclusive for a test that is supposed to be guarding against claims that monopolize abstract ideas. Second, it’s not much of a test at all if it’s a foregone conclusion for computer-related inventions. Alice Step 2 thus often devolves into a mere formality that is always decided the same way. The result is that if a computer-related claim fails Alice Step 1, you lose. There is, in practice, no second chance to demonstrate eligibility.
Thus, Alice Step 2 at the USPTO has recently become a vestigial remnant that poses great danger for modern AI inventions. This shift toward a predetermined version of Alice Step 2 was seemingly a response to the burdens imposed by the Berkheimer Memo, which required onerous fact-finding for examiners to sustain eligibility rejections.
As one stark data point, . This is remarkable not just for the 0% success rate, or for the 20,000 words in the Manual of Patent Examining Procedure about how to find an inventive concept, but also because it’s a change from what the Office did before Berkheimer, when the PTAB routinely held that claims contained an inventive concept under Alice Step 2.
Some clarification from the Federal Circuit would be welcome. The inadequacy of conventional computers has vast support in the Federal Circuit case law. One of the bedrock principles of eligibility is that merely reciting a conventional computer is insufficient to supply an inventive concept under Alice Step 2. But it’s a logical error to turn that observation into the entire test.
A Better Way
I’ve long thought that when Federal Circuit cases use the term “conventional,” they do not exactly mean “old,” but rather something like “operating in an expected way.” In that sense, being conventional by operating in an expected way meshes with the Alice Step 2 inquiry by essentially asking, “are these features, operating in their ordinary and expected way, merely a drafting effort that attempts to monopolize the abstract idea?”
It’s for this reason that I have been waiting for the Federal Circuit to use the missing word “inherent.” That is, claim features that are inherent to practicing the abstract idea are insufficient to provide an inventive concept. This is a powerful tool of inquiry because it efficiently cuts through patent attorney fluff when it is a mere attempt to dress up an ineligible claim.
This term also has a strong policy resonance with the Patent Act because features that are inherent to a process are features that would be immediately apparent to a skilled artisan. And, therefore, the state of the art is not advanced by the disclosure of such features, and the public gets nothing in return for the grant of a patent monopoly on inherent features.
There have been prior cases that used similar reasoning. For example, in Boom! Payments, Inc. v. Stripe, Inc., a nonprecedential opinion decided in January of 2021, Judge Lourie wrote that the purported inventive concepts advanced by the patentee were merely the “necessary steps” of practicing the abstract idea. Slip op. at 9. In a little-publicized, nonprecedential opinion earlier this year, Plotagraph Inc. v. Lightricks, Ltd., Judge Schall wrote that each of the patentee’s purported inventive concepts were “inherent in nonautomated computer animation.” Slip op. at 11. All the pieces seem to be in place for the Federal Circuit to clarify that Alice Step 2 is not a search for generic computers.
Clarification is Essential
The USPTO is currently . As it does so, it should strongly consider clarifying that if claims are to be found ineligible under Alice Step 2, it should not be merely because they recite computers that are generic or conventional, but rather because they include only features that are inherent to monopolizing an abstract idea—features that cannot supply an inventive concept.
Restoration of Alice Step 2 is essential for our patent system to ensure that the most important AI inventions are afforded sufficient protection.
