Legal Alert: USPTO Updates Eligibility Guidance on AI Inventions

On July 17, the USPTO published an update to the patent eligibility guidance and added three new examples to aid practitioners and examiners in determining whether a claimed invention is eligible subject matter under Section 101. The updated guidance addresses several points on how to assess the eligibility of AI and machine learning inventions specifically.

In general, the updated guidance is an incremental change that retains the previous overall eligibility determination framework having a Step 2A, Prong 1 for determining whether the claim recites a judicial exception, and a Step 2A, Prong 2 for determining whether a recited judicial exception is integrated into a practical application.

The primary takeaway from the new guidance is that AI inventions are most likely to be deemed eligible when they solve a particular technological problem or improve another technology. Importantly, the advance described in the specification must be reflected in the claims to demonstrate eligibility.

Under Step 2A, Prong 1, the updated guidance indicates that AI inventions are at risk of reciting a judicial exception in three primary ways: 1) for reciting mathematical calculations, 2) for reading on a mental process, and 3) for reciting merely the idea of a solution rather than details about how a solution is implemented.

The updated examples include a hypothetical claim that is ineligible because it merely recites the mathematical calculations for generating embedding vectors and without reflecting the disclosed improvement. The hypothetical claim thus fails to recite how the solution is accomplished or how the claim provides a technological advantage.

For mental processes, the updated examples include a hypothetical claim that recites using a trained neural network to detect anomalies, but only at a high level. The claim therefore reads on the mental process of detecting anomalies, with the trained neural network being interpreted as a mere field of use.

Under Step 2A, Prong 2, the updated guidance makes a few important clarifications. Namely, the guidance clarifies the role of the so-called “additional elements” recited beyond a judicial exception. In particular, the guidance states that the additional elements need not be the only features that provide an improvement or technological solution. Instead, the additional elements, in combination with any recited judicial exception, can demonstrate integration into a practical application. In addition, the guidance states that under Step 2A, Prong 2, it is immaterial whether the additional elements are well understood, routine, and conventional. These clarifications provide practitioners with useful tools to demonstrate the eligibility of their claims for AI inventions.

Lastly, the updated guidance and examples appear to reconfirm the diminishing role of Step 2B in the eligibility analysis. The updated guidance and examples make no significant changes to the way the Office interprets Step 2B, and they do not provide any example of a claim that is ineligible under Step 2A but eligible under Step 2B. Thus, successful arguments regarding Step 2B would seem to be limited only to procedural defects, such as an examiner skipping the analysis entirely.

In conclusion, the updated guidance reaffirms the conventional wisdom that claims which recite black box AI or machine learning modules may be viewed skeptically under the USPTO’s eligibility analysis. To reduce the risk of facing such problems, practitioners should ensure that their specifications articulate an improvement over the prior art and that their claims reflect those improvements.

Read the guidance in the Federal Register and the examples on USPTO.gov.