Meta Hit with Class Action Over AI Training on Facebook and Instagram Photos

A new proposed class action filed against Meta alleges the company illegally harvested photos from Facebook and Instagram to train its AI image-generation models and to develop a face-recognition feature called NameTag that has not been released publicly. The complaint, first reported by Wired, claims Meta used the images without obtaining proper consent from the people who posted them. The lawsuit seeks class-action status on behalf of users whose photos were allegedly swept into the training pipeline.

The core of the allegation is that Meta treated its vast trove of user-uploaded photos as a free training resource for two separate AI projects. One is the company's image-generation technology, which competes with tools from OpenAI, Google, and others. The other is NameTag, a face-recognition system that Meta built but never launched, according to the suit. The plaintiffs argue that users never agreed to either use when they uploaded pictures to the platforms.

Face recognition has been a legal minefield for Meta before. The company paid $650 million in 2021 to settle a class action in Illinois under the state's Biometric Information Privacy Act over its earlier photo-tagging feature, which automatically identified people in uploaded images. That case, Patel v. Facebook, became a landmark for biometric privacy law. The new complaint suggests Meta may have continued building similar technology even after that settlement, though the company has not commented on the specifics.

The lawsuit also arrives as Meta faces growing scrutiny over how it sources data for AI. In recent years, the company has acknowledged scraping public web data for training, and it has updated privacy policies to allow AI development using user content. But the plaintiffs in this case draw a line between public posts and the private, granular use of personal photos to build facial-recognition models. If the court agrees, it could force Meta to overhaul how it collects and uses image data across its apps.

What the suit does not yet establish is whether Meta actually used every user's photos or only a subset, and whether NameTag ever reached a functional prototype. Meta has consistently said its AI training practices comply with applicable laws, and it has not responded directly to this complaint. The case will likely hinge on Meta's terms of service and whether they adequately disclose AI training and biometric processing to ordinary users.

For the broader AI industry, the outcome could set a precedent for how companies use social media images to train generative models. Other platforms with large photo libraries, including Google and TikTok, face similar questions about consent and transparency. If the court certifies the class, Meta could be forced to disclose internal documents about its training data and the fate of NameTag. The next key step is Meta's response to the complaint, expected in the coming weeks.