Photo of John Gray

Judge William Orrick of the Northern District of California held a hearing last week to address 12(b)(6) motions to dismiss by Facebook and Princeton University in UAB “Planner5D” v. Facebook, Inc., et al., Case No. 19-CV-03132 (N.D. Cal.). This appears to be one of the first trade secret misappropriation and copyright cases relating to artificial intelligence datasets.

Plaintiff Planner 5D has a website with a home design tool to allow customers to digitally design their own home, office, or landscape. As alleged by Planner 5D, the tool uses thousands of object and scene files as seed data sets to train machine learning algorithms. The complaint alleges that Princeton used software tools to obtain secret internet addresses where the object and scene files were hidden and then scraped the website to obtain the files, ultimately making the data available to Princeton researchers. Princeton then allegedly shared that data with Facebook. Planner 5D contends that, although the images themselves are viewable on the internet, the underlying datafiles and secret internet addresses that were scraped by Princeton, as well as the file locations, are trade secrets.

This is the second motion to dismiss. (The Court previously granted Defendants’ motions to dismiss because the complaint did not allege sufficiently specific facts on the trade secret claim and because the copyrights were not registered.) Defendants’ second motion to dismiss argues that Planner 5D made its materials available to the public on the internet—meaning the information could not qualify as a trade secret—and that Planner 5D’s Terms of Service do not require confidentiality. Planner 5D counters that the secret URLs were nonpublic and that the “https” provided a secure encrypted site. But Defendants point out that the “secret” URLs were not password protected and did not have other forms of defense to prevent access. Indeed, the materials could be accessed through what Defendants described as standard software tools. As to the claims specifically against Facebook, Defendant Facebook argues that the allegations are conclusory and that Planner 5D failed to allege a link between Princeton’s scraping and Facebook because Facebook’s mere possession of alleged trade secret information, without more, is not misappropriation.

The Court took the ruling under advisement. The order could be interesting to watch for potential rulings relating to whether and when information on the internet can constitute a trade secret, whether scraping can constitute misappropriation, and whether terms of use for a website can be used to show either reasonable efforts to protect a trade secret or misappropriation if the information is gained in violation of the terms of use.