IP Organizations Want More on USPTO’s AI Patent Eligibility Guidance

The deadline for comments on the U.S. Patent and Trademark Office’s (USPTO) updated subject matter eligibility guidance was October 16, and the Office received 24 total submissions. The 19 posted thus far overwhelmingly call for more detail in the guidance in order to avoid undue restrictions on patentability of critical artificial intelligence (AI) technologies.

Major IP organizations, including the American Intellectual Property Law Association (AIPLA), the Council for Innovation Promotion (C4IP), the Intellectual Property Owners Association (IPO), the Pharmaceutical Research and Manufacturers of America (PhRMA) and the Association of University Technology Managers, Inc. (AUTM), have now weighed in. Many of them seem concerned that USPTO Director Kathi Vidal’s July 2024 update to the Office’s Subject Matter Eligibility (SME) Guidance includes potentially conflicting examples and a lack of detailed explanation.

Vidal announced the SME update on July 16 of this year in an effort to more directly address inventions involving AI, particularly with respect to “(1) the evaluation of whether a claim recites an abstract idea in Step 2A, Prong One; and (2) the evaluation of the improvements consideration in Step 2A, Prong Two.” The updated guidance includes three new and detailed examples using hypothetical claims to address common situations, such as “whether a claim recites an abstract idea or whether a claim integrates the abstract idea into a practical application.”

At the time of the announcement, Vidal said that while the Office has “a lot of certainty behind” the guidance, which became effective as of July 17, the USPTO will consider the comments received. “The best feedback is the most concrete feedback,” Vidal told IPWatchdog. “Examiners will also have an opportunity to give feedback – they’re going to get training and if they think there’s something that can be better said, they can give us feedback.”

AIPLA: Eligibility Guidance Needs More Clarity on Applying Step 2 of Alice/Mayo

More detailed guidance on the proper application of step two of the Alice/Mayo test was a focal point of AIPLA’s comment . The national IP bar association argues that a lack of guidance for Alice/Mayo Step 2A, Prong One, which evaluates whether a patent claim is directed to a judicial exception, leads to a potential conflict in SME examples published by the USPTO. For example, new Example 47 includes an application of AI model training that is directed to an abstract idea despite being more narrowly claimed than Example 39’s neural network training, which is patent-eligible in that example. AIPLA also calls for more guidance on distinguishing claims directed to judicial exceptions at Step 2A, Prong Two, and for the agency to meaningfully address Step 2B analysis to identify specific improvements to computational processes.

AIPLA’s comment also includes several specific concerns with Examples 47 through 49 published by the USPTO for SME guidance in the AI context. Not only are these three examples not representative of the field of AI inventions claimed in patent applications filed at the USPTO, but AIPLA also contends that each example relies on a problematic claim construction conflating eligibility analysis with interpretation.

C4IP: The USPTO Needs to Advocate for Legislative Reform to Section 101

The USPTO’s eligibility guidance from this July pays short shrift to several technical considerations important to AI patent-eligibility, according to C4IP’s comment. For instance, the Office’s treatment of trained artificial neural network (ANN) claims in steps (d) and (e) of Example 47 directs examiners to treat the trained ANN as a generic computer without any support for that proposition. The July 2024 guidance’s summary of relevant Federal Circuit case law is also problematic for describing those cases in a way that invites examiners to ignore concrete claim limitations, C4IP contends.

C4IP further points out that the non-binding nature of USPTO guidance on courts contributes to a false sense of security for AI patent applicants. These potential validity issues threaten investment in a technological field critical to American competitiveness, C4IP argues. The coalition of former judges and government officials ended by urging the USPTO to take a proactive role in advocating for legislative reform on Section 101, including passage of the Patent Eligibility Restoration Act (PERA).

PhRMA: Train Patent Examiners to Recognize Eligibility of AI-Assisted Inventions

PhRMA, which represents biopharmaceutical companies, underscored its agreement with the USPTO’s July 2024 statements that the use of AI to assist in invention development is not a consideration under the Alice/Mayo test. The biopharma association’s comment  urged the USPTO to train patent examiners on this point to properly focus the SME analysis on a patent’s claims.

While PhRMA appreciated the usefulness of the USPTO’s new SME examples in the AI context, the organization argued that some of the agency’s concerns were more properly dealt with under 35 U.S.C. § 112 rather than Section 101. Further, given similarities between new Example 49’s method of treating fibrosis, which has an identifying step assisted by AI, and Example 29’s method of treating the fictional autoimmune disease “julitis,” PhRMA suggested that the USPTO clarify that Example 49 does not overwrite the Example 29 guidance.

IPO: Non-Example-Based Guidance on Patent-Eligible Inventions Would Be Useful

Like AIPLA, the comment filed by IPO asks the USPTO for further clarity on applying both prongs of Step 2 of Alice/Mayo. Along with additional examples of AI inventions that are patent-eligible at Step 2, Prong One, the international IP owners’ association also asked the USPTO to develop non-example-based guidance for distinguishing high-level mental processes and mathematical algorithms from patent claims sufficiently detailed for patent eligibility.

IPO also questioned the agency’s treatment of certain claim elements in the new AI invention examples as “insignificant extra-solution activity.” Whereas elements (a) and (f) from Claim 2 of Example 47’s trained ANN are deemed to be insignificant for eligibility, IPO notes that (a)’s receiving of continuous training data allows the deployment of the invention in a real-world application and prevents preemption that could be caused by other claim element combinations.

AUTM: Section 101 Ineligibility Should Be the Same 50 Years Ago and 50 Years In Future

Noting that SME analyses often import patentability considerations outside of Section 101, AUTM’s comment notes that the patent-eligibility determination should not vary with time. “If an invention is ineligible for patent protection now, the same should have been true 50 years ago and should be true 50 years from now,” AUTM writes. The academic research institution organization also cautioned the USPTO against rules that automatically consider claims containing mathematical formalisms to be directed to abstract ideas.

Along with more SME examples demonstrating the smallest necessary change to meet the “significantly more” threshold of Alice/Mayo Step 2B, AUTM asks the USPTO to reconsider its eligibility analysis in the new examples, especially for data processing and nonobvious uses of known hardware. AUTM requests specific new examples related to multiple robot scheduling activity, high throughput screening using machine learning and method of alerting users to suspicious Internet domains.

Image Source: Deposit Photos
Image ID:110249524
Copyright:kikkerdirk