The patentability of artificial intelligence (AI) has been increasingly scrutinized in light of the surge in AI technology development and the ambiguity regarding the interpretation of software-related patents. The Federal Circuit has gradually refined the criteria for determining subject matter eligibility for software-related patents, and based in part on such jurisprudence, earlier this year the U.S. Patent and Trademark Office (USPTO) released revised guidance on examining patent subject matter eligibility under 35 U.S.C. §101. See 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (Jan. 7, 2019). Considering the advances in AI technology and intellectual property law, how do these recent developments shape the outlook of AI patentability?

Current USPTO Treatment of AI

The USPTO defines “AI” as including artificial intelligence type computers, digital data processing systems, corresponding data processing methods, products for emulation of intelligence, “(i.e., knowledge based systems, reasoning systems, and knowledge acquisition systems),” systems for reasoning with uncertainty, adaptive systems, machine learning systems, and artificial neural networks. U.S. Patent and Trademark Office, Class 706, Data Processing—Artificial Intelligence (January 2011). However, AI technologies may also be assigned to one of dozens of other software-related classes, and if so, patent applications on such technologies may have varied outcomes.

For example, Class 706 encompasses art unit 2121, which examines patent applications related to “Data Processing: Generic Control Systems or Specific Applications,” and art unit 2129, which examines patent applications related to “Data Processing: Artificial Intelligence.” U.S. Patent and Trademark Office, Classes Arranged by Art Unit: Art Units 1764-2691 (last visited Feb. 19, 2019). Art units 2121 and 2129 are assigned to different supervisory patent examiners, and according to LexisNexis PatentAdvisor, art unit 2121 has an allowance rate of 72.7 percent, while art unit 2129 has an allowance rate of 80.4 percent. U.S. Patent and Trademark Office, TC 2100 Management Roster (last visited Feb. 19, 2019). Since an application's assigned art unit may depend on the details of the technology and wording of the specification and claims, the potential for different allowance rates highlights one of the reasons that careful drafting of AI patent applications is crucial.

Moreover, as noted by legal commentators in a recent USPTO conference on AI, the USPTO may be paying more attention to AI patents pursuant to the Federal Circuit's interpretation of Electric Power Group v. Alstom S.A., 830 F.3d 1350 (Fed. Cir. 2016). See Artificial Intelligence: Intellectual Property Policy Considerations, Conference at the U.S. Patent and Trademark Office (Jan. 31, 2019). In Electric Power Group, claims directed to systems and methods for monitoring the performance of an electric power grid were held ineligible under §101 as a result of “merely requiring the selection and manipulation of information.” Id. at 1354-55. The Federal Circuit noted that “we have treated analyzing information by steps people go through in their minds, or by mathematical algorithms, without more, as essentially mental processes within the abstract-idea category.” Id. at 1354. If the USPTO follows an Electric Power Group interpretation of an abstract idea, the careful wording of AI patent applications may be increasingly important in ensuring such applications are allowed. Indeed, recent case law interpreting §101 draws attention to the significance of thoughtful drafting of AI patents.

Recent Court Decisions

More recent cases have underscored the legal scrutiny of patents relating to AI technologies. Last year, in Finjan v. Blue Coat Systems, 879 F.3d 1299, 1305-06 (Fed. Cir. 2018), the Federal Circuit found that claims directed to a method of virus scanning were patent-eligible under Step 2A of the USPTO's subject matter eligibility test. Conversely, in a district court case pertaining to AI, PurePredictive v. H2O.AI, No. 17-cv-03049-WHO, 2017 WL 3721480, at *1 (N.D. Cal. Aug. 29, 2017), the Northern District of California found that a claim for a machine learning predictive analysis framework was directed to “the abstract concept of the manipulation of mathematical functions and make[s] use of computers only as tools, rather than provid[ing] a specific improvement to computer related technology.” The Federal Circuit affirmed this decision without opinion in November 2018. PurePredictive v. H2O.AI, 741 F. App'x 802 (Fed. Cir. 2018). These decisions may present further limitations on the patentability of certain AI technologies.

Revised USPTO Guidance

However, the recently released USPTO guidance on subject-matter eligibility may offer hope for potential applicants of AI patents. In January 2019, the USPTO released revised subject matter eligibility guidance to clarify its interpretation of the Alice/Mayo test. See 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (Jan. 7, 2019); see also Alice Corp. Pty. Ltd. v. CLS Bank Int'l, 573 U.S. 208, 217-18 (2014) (citing Mayo Collaborative Servs. v. Prometheus Labs., 566 U.S. 66 (2012)). This revision was released in part due to a “need for more clarity and predictability” of the USPTO's patent subject matter eligibility test as applied to subject matter falling within a judicial exception. 84 Fed. Reg. at 50. To address this and other concerns, the USPTO modified Step 2A of the Subject Matter Eligibility Guidance as incorporated into the Manual of Patent Examining Procedure. Id.

The USPTO revised Step 2A by (1) providing groupings of subject matter considered abstract ideas and (2) instructing that “a claim is not 'directed to' a judicial exception if the judicial exception is integrated into a practical application of that exception.” Id. Specifically, the first prong of Step 2A determines whether a claim is directed to a judicial exception (laws of nature, natural phenomena, and abstract ideas). Id. at 52. The USPTO limited abstract ideas to the following groupings: mathematical concepts, methods of organizing human activity, and mental processes. Id. If the claim is directed to one of these judicial exceptions, the claim must be examined under the second prong of Step 2A, which considers “whether the claim recites additional elements that integrate the exception into a practical application of that exception.” Id. at 54. If so, then the claim is patent-eligible subject matter, but if not, the claim must be examined under Step 2B, which considers whether the claim is directed to an inventive concept. Id. at 56.

The revised §101 guidance may be good news for potential applicants and assignees of AI patents. The USPTO's guidance narrows the scope of the judicial exceptions by limiting abstract idea subject matter to specific groupings, so it is possible that a particular AI technology may not fall into any of the categories if many of the necessary recited steps cannot be practically performed or applied in the human mind due the amount of processor power required or data to be analyzed. For example, the USPTO provides an example claim directed to “[a] computer-implemented method of training a neural network for facial detection.” U.S. Patent and Trademark Office, Subject Matter Eligibility Examples: Abstract Ideas 8-9 (Jan. 7, 2019). The USPTO explains that this claim does not recite any of the judicial exceptions. First, the claim does not recite any mathematical relationships, functions, or calculations and “[w]hile some of the limitations may be based on mathematical concepts, the mathematical concepts are not recited in the claims.” Id. at 9. Second, the claim does not recite a mental process because the recited steps required to train a neural network are “not practically performed in the human mind.” Id. Third, the claim “does not recite any method of organizing human activity.” Id. Thus, under the revised Step 2A, the claim recites eligible subject matter. Id.

Additionally, under the revised Step 2A, a claim may be considered “integrated into a practical exception” depending on the claim language and arguments used. For example, the USPTO provides another sample claim directed to “[a] method for adaptive monitoring of traffic data through a network appliance connected between computing devices in a network.” Id. at 10-11. The USPTO explains that this claim would be patentable under Step 2A because the “claim as a whole integrates the mental process into a practical application” and “the claim … is directed to a particular improvement in collecting traffic data,” resulting in “improved network monitoring.” Id. at 11. Thus, the patentability of certain AI technologies under the revised guidelines may still rely on the details of how the claims are drafted.

Conclusion

For now, both the revised USPTO guidance and the recent Federal Circuit jurisprudence highlight the increased importance of strong arguments and thoughtful claim drafting in distinguishing between an unpatentable abstract idea and a patentable improvement to a technology. Those wishing to patent AI technologies may want to watch carefully to see how the USPTO and Federal Circuit interpret the new guidance.

Jennifer Tempesta is a partner and Stephanie Kato is an associate in the New York intellectual property group at Baker Botts.