The theory is that Meta took faces out of Facebook and Instagram, converted them into biometric identifiers, and used those identifiers to train both its image generation models and the face recognition system it had been preparing for its AI glasses. The proposed class period opens on September 4, 2021.
The exhibit that came out of the patent office
Paragraphs 99 through 105 of the complaint reach past the reporting and into Meta’s own patent filings.
United States patent publication 20260142974 A1 is titled “User identity verification without sharing biometric data with platforms.” Meta Platforms is the applicant. The named inventor is Shing Sheung Daniel Ip. It was filed on October 9, 2025 and published on May 21, 2026, roughly a month before Meta Glasses went on sale. The document describes a device capturing an image or video of a face and generating a biometric face embedding from it. That embedding is then compared against embeddings derived from profile photographs and tagged photographs the platform already holds. Smart glasses are among the device types it lists.
The plaintiffs call this a blueprint. Paragraph 104 argues the application “does not describe a hypothetical or generic computer-vision concept” but a complete operational workflow. Paragraph 105 concludes that Meta “had designed and implemented the core technical components necessary to identify real-world individuals from images captured by Meta smart glasses.”
Then the complaint concedes the weakness in its own exhibit, which is not a common move. Paragraph 103 acknowledges that the patent “emphasizes a privacy-preserving architecture in which certain biometric embeddings need not be transmitted from the user’s device to Meta’s servers.” The title says as much in its last six words. A patent describing how to verify somebody without handing biometric data to a platform is an awkward thing to cite as evidence that a platform is hoarding biometric data, and the plaintiffs clearly know it. Their answer is that the filing proves capability rather than practice. The pipeline exists; what Meta chose to do with the output is a separate question. That is a serviceable argument, and it is a narrower one than the case looks like from the headlines.
Where NameTag came from
NameTag is the dormant face recognition code that WIRED reported finding in the Meta AI companion app in June. The complaint quotes that reporting at paragraph 22. The feature would turn faces captured by the glasses into “unique biometric signatures, commonly known as faceprints, and check each one against faceprints stored on the user’s phone.” That phone-side database, WIRED found, was configured to receive updates from Meta.
The complaint describes NameTag as code that reached phones without ever being switched on, and alleges at paragraph 45 that Meta has since removed it from the app. Removal settles nothing, the plaintiffs say, because the identifiers the feature would have matched against were created and stored long before any of it shipped.
