Meta Sued Over Data Used to Train AI and Build Face-Recognition System

Summary

Parents and children in Illinois and California have filed a proposed federal class action accusing Meta of using Facebook and Instagram photos without proper notice or consent. The complaint alleges that the images supplied biometric information for NameTag, an unreleased face-recognition feature designed for Meta’s smart glasses, and for generative image systems including Emu and Muse Image. According to the complaint, NameTag code was embedded in Meta’s glasses AI companion app, which had been downloaded more than 50 million times, although the feature was not enabled for consumers. WIRED’s analysis and outside testing found that the code could convert faces captured by the glasses into biometric signatures and compare them with faceprints stored on a user’s phone; the complaint says the source of those faceprints remains known only to Meta. The filing cites reporting about possible recognition through Meta connections or public Instagram accounts, as well as a Meta patent describing face matching against profile photos and other company-held images. Meta says the lawsuit lacks merit, maintains that it has been transparent about its information practices, and says NameTag has not shipped and that no final launch decision has been made. The company also says it is not building a universal face database. The proposed class could include people in Illinois, California, and elsewhere in the United States whose images appeared on the platforms or were submitted to Meta’s generative AI systems through prompts since September 4, 2021; the complaint estimates millions of potential members. The plaintiffs seek statutory and actual damages under Illinois and California law, along with injunctive relief. The case follows Meta’s $650 million Illinois biometric settlement in 2020, its 2021 shutdown of an earlier face-recognition system and deletion of more than a billion faceprints, and a $1.4 billion Texas biometric-data settlement in 2024. Meta removed the NameTag code the day after WIRED’s June report, while continuing to describe the feature as exploratory.