Artificial intelligence (AI) has recently exploded onto the scene as a technology that will redefine the next technological age. It is estimated that 38 percent of all businesses are implementing AI in one form or another, and that adoption rate is expected to grow to 62 percent in 2018. See “Artificial Intelligence Collides with Patent Law”, World Economic Forum White Paper (April 2018). Similarly, global revenue from AI-based systems is expected to grow from nearly $8 billion in 2016 to more than $47 billion in 2020. See Bloomberg Law, “2017 Outlook IP, Privacy, Tech and Telecom,” Daily Report for Executives (2016).
The rise in capital investment on AI-based technologies has driven a similar increase in global intellectual property rights investment. From 2012 to 2017, the USPTO saw a 500 percent increase in the number of patents issuing to class 706—a classification exclusively designated for AI. See “Intellectual Property Protection for Artificial Intelligence”, Westlaw Journal Intellectual Property (Aug. 30, 2017).
The emergence of AI, however, has coincided with uncertainty of the patentability framework for protecting such innovations. AI, which embodies processes and machines, fits comfortably within the statutory definition of patentable subject matter. See 35 U.S.C. §101. However, the Supreme Court’s decision in Alice Corp. Pty. Ltd. v. CLS Bank Int’l created additional hurdles for software and “computer-implemented inventions,” such as AI. Specifically, Alice created a two part test: (1) whether the claims embodying the invention are directed to an “abstract idea” (known as Step 2A); and if so, (2) whether or not the claims recite significantly more than the abstract idea (known as Step 2B). 134 S. Ct. 2347, 2354-55 (2014).
Fortunately, the Federal Circuit issued a collection of opinions that provide guidance for protecting AI-based inventions. And the new director of the USPTO has opined on the importance of subject matter reform. See April 11, 2018, Remarks by Director Andrei Iancu at U.S. Chamber of Commerce Patent Policy Conference. These recent trends in case law and policy coupled with the right strategies in patent preparation and prosecution provide a discernable path to patentability for AI-based inventions.
Post-’Alice’ Eligibility Guideposts
Practitioners looking to protect AI-based inventions should start with a study of the Federal Circuit cases finding software and computer-implemented innovations patent-eligible. The reading list includes: DDR Holdings v. Hotels.com; Enfish v. Microsoft Corp.; BASCOM Global Internet Servs. v. AT&T Mobility; McRO v. Bandai Namco Games Am.; Amdocs (Isr.) Ltd. v. Openet Telecom; Trading Techs. Int’l v. CQG; Thales Visionix v. United States; Visual Memory v. NVIDIA Corp.; Finjan v. Blue Coat Systems; and Core Wireless Licensing S.A.R.L. v. LG Electronics.
Enfish, Bascom, McRo and Finjan are particularly relevant to AI-based inventions.
Enfish, which found claims directed to an improved database architecture eligible under Step 2A of the Alice framework, provides the foundational principle that “[s]oftware can make non-abstract improvements to computer technology just as hardware improvements can.” Enfish v. Microsoft Corp., 822 F.3d 1327, 1335 (Fed. Cir. 2016). Thus, Enfish should be applicable in situations in which an AI-based invention improves a computer-based technology, such as, self-driving car technologies.
Bascom found claims directed to a system for customizable filtering of Internet content eligible under Step 2B of the Alice framework. BASCOM Global Internet Servs. v. AT&T Mobility, 827 F.3d 1341, 1349-50 (Fed. Cir. 2016). There, the Federal Circuit noted that “an inventive concept can be found in the non-conventional and non-generic arrangement of known, conventional pieces.” Id. at 1350. This reasoning is particularly relevant to the models underlying AI-based inventions because while the types of models may be generally known, the “non-conventional and non-generic” configuration of such models may be nevertheless patentable.
McRo found claims directed to automation of animation tasks eligible under Step 2A of the Alice framework. McRO v. Bandai Namco Games Am., 837 F.3d 1299, 1316 (Fed. Cir. 2016). There, the Federal Circuit noted that “[t]he claimed process uses a combined order of specific rules that renders information into a specific format that is then used and applied to create desired results.” Id. at 1315. AI may be based on similar sets of “specific rules” that likewise “render information into a specific format that is then used and applied to create desired results,” and so should be patentable under this same reasoning.
Finally, Finjan found claims directed to a method of virus scanning eligible under Step 2A of the Alice framework. Finjan v. Blue Coat Systems, 879 F.3d 1299, 1305-06 (Fed. Cir. 2018). There, the Federal Circuit noted that the claims recite specific steps that accomplish a result that realizes an improvement in computer functionality, id., which seems to combine the reasoning of Enfish and McRo. Because virus scanning is a classic AI-type technology, the reasoning in Finjan should be applicable to other AI-based innovations.
AI Patent Application Eligibility Strategies
The legal guideposts described above provide opportunities for maximizing AI patent protection. In particular, the cases decided since Alice have consistently highlighted the essential nature of the specification in the subject matter inquiry, and this realization can be exploited during the preparation and prosecution of AI applications.
Specifically, for new applications, addressing in detail the technical problem and the technical solution is critical. Fortunately, Europe has long used this approach, so not only is there a great body of examples, but this strategy may improve how those cases fare when later entering Europe. Though this recommendation is broadly applicable to software-based innovations, it is particularly relevant to AI-based innovations to help overcome the Step 2B “significantly more” hurdle.
Further, when describing the invention, it is crucial to be means-oriented, not only ends-oriented. Avoid the “black box” logic of input data, process data, output data. Federal Circuit cases such as Elec. Power Grp. v. Alstom, which held that “merely selecting information, by content or source, for collection, analysis, and display does nothing significant to differentiate a process from ordinary mental processes,” are simply too broad in their sweep to rely on the input and the output plus an inference of what happens in the middle. 830 F.3d 1350, 1355 (Fed. Cir. 2016); but see Ex Parte Faith, Appeal 2016-008020, Application 12/778,638 (PTAB March 30, 2018) (“We recognize that ‘collecting information, including when limited to particular content (which does not change its character as information), [is] within the realm of abstract ideas.’ However … the claims before us are different as they change the character of information via the claimed encryption operations.”) (emphasis in original). Instead, provide practical implementation details, such as specifics of the model design. Relatedly, draft claim elements that recite specific uses of resulting data, not just the generation of the data. To these ends, it is incumbent upon practitioners to extract more detail in invention disclosure meetings.
It is also important to include reasoning and examples in the application that mimic the content specifically relied upon in the case law finding claims subject matter eligible. In other words—stack the deck in the applicant’s favor.
Applications written post-Alice and utilizing these recommendations will fare far better in subject matter eligibility challenges. But there is still a large backlog of applications written pre-Alice that require attention.
Fortunately, the Federal Circuit recently gifted applicants with Berkheimer v. HP Inc., 881 F.3d 1360, 1365 (Fed. Cir. 2018). Berkheimer provides two critical tools for overcoming subject matter eligibility rejections.
First, Berkheimer is unequivocal that all claims must be considered individually under Section 101. Berkheimer, 881 F.3d at 1365 (“A claim is not representative simply because it is an independent claim.”). This reasoning can be used to overcome the common “form” rejections wherein an examiner analyzes one independent claim and then states categorically that the same reasoning applies to all claims. Berkheimer-based responses have proven very effective to overcome this short-circuit of a proper analysis.
Second, Berkheimer states that examiners cannot argue something is “routine, well-known, or conventional” without providing “clear and convincing evidence.” Berkheimer, 881 F.3d at 1368 (“[t]he question of whether a claim element or combination of elements is well-understood, routine and conventional to a skilled artisan in the relevant field is a question of fact” that must be “proven by clear and convincing evidence.”); see also Aatrix Software v. Green Shades Software, 882 F.3d 1121, 1129 (Fed. Cir. 2018) (Examiners cannot take notice of conventionality adverse to an applicant). Moreover, the Federal Circuit made clear that “[t]he mere fact that something is disclosed in a piece of prior art … does not mean it was well-understood, routine, and conventional.” Id.; see also Exergen Corp. v. Kaz USA, 725 F. App’x 959, 965 (Fed. Cir. 2018) (same). This is useful to overcome another frequent deficiency in subject matter eligibility rejections.
Finally, examiner interviews have never been more important. Applicants can get useful feedback from examiners during interviews regarding claim amendments to overcome subject matter eligibility rejections, which would never show up in a written rejection. In some cases, an interview can make clear that more responses would be futile, and give applicants a chance to advance to appeal before reaching their RCE pain threshold.
AI is already changing, and will continue to fundamentally change the technology landscape. Some are already referring to AI as the “Fourth Industrial Revolution.” Fortunately, despite the frustrations of Alice, clear strategies have emerged for protecting AI-based inventions. So like the revolutionaries that have come before and benefited from continual investment in the protection of their intellectual property, the leading innovators of this coming revolution should continue to pursue protection of their AI-based intellectual property.
Nick Transier is a registered patent attorney with Patterson + Sheridan, whose practice includes preparation and prosecution of software and electronics-based patent applications, including AI and machine learning. Keith Toboada is managing partner of the Shrewsbury office of Patterson + Sheridan.