As artificial intelligence becomes increasingly integral to innovation in many industries, the patent eligibility of machine learning inventions remains uncertain. However, recent IP law developments have shed some light on this issue, addressing whether training an AI model is enough to constitute a patentable product.

Panitch Schwarze attorney Sean M. Douglass recently authored an article in The Legal Intelligencer examining the criteria used to establish patentability in these cases. In Recentive Analytics v. Fox, the U.S. Court of Appeals for the Federal Circuit found that simply training a machine learning model does not transform an abstract idea into something that is patent eligible. The United States Patent and Trademark Office (USPTO) has provided further clarification with recent patent eligibility examples. This guidance indicates that generic machine learning operations, claimed at a high level or expressed as mathematical concepts, are too abstract to be considered eligible; however, machine learning that is applied to a concrete technological advance can be patentable.

When drafting patent applications, applicants should be careful to highlight the technical solutions to specific problems offered by machine learning inventions. These details can make the difference when seeking to protect valuable AI innovations.

Read the full article here: Training Alone Is Not Enough: Lessons From ‘Recentive’ and USPTO AI Examples on Patent Eligible Machine-Learning Claims (Subscription is required.)