Client H2O Secures Important Federal Circuit Win in Clash Over AI Technology


The Federal Circuit has rejected patent infringement claims against our client H2O in a decision that backed our IP and appellate team’s arguments and secures critical protections for the company’s groundbreaking open-source machine learning platform.

PurePredictive had asserted patent infringement claims in the Northern District of California based on its ’446 patent, which describes a method of performing predictive analytics – using data about past events to predict future events – by pseudo-randomly generating potential algorithms, evaluating those algorithms and determining which of those should be used for certain subsets of data.

The claims posed a threat to H2O’s platform, which is used by thousands of organizations globally to power a wide range of applications. Orrick partner Clem Roberts, with his former team, won dismissal of the claims in the lower court on the grounds that PurePredictive’s patent covered nothing more than an abstract idea and therefore is not patent eligible. The Federal Circuit upheld that dismissal on appeal, which was argued by partner Mel Bostwick. Mel teamed with Clem on the case.

The decision is considered important in the current debate over the question of when a patent claims an abstract idea and whether it’s proper to decide that issue at the Rule 12(b) (6) stage. By affirming the lower court dismissal, the Federal Circuit rejected PurePredictive’s effort to argue that its patent actually covered a particular data structure and that the district court should have conducted claim construction or considered factual disputes.

Instead, the outcome supported Orrick’s argument that there was no structure to this invention, as it could be used in any application and any format, and that it was just the manipulation of data untethered to anything specific. The decision provides important support for the concept that patents in the machine learning/AI space must be directed to specific technological improvements.

"Machine learning tools like H2O’s Auto ML are critically important to an extremely wide range of applications and the court correctly recognized that patents in this area have to be directed to specific technological improvements and not merely to the idea of automatically generating and refining algorithms. I give H2O a lot of credit for firmly standing on principle and seeing this through to the very end," Mel told Law360.