Meta Sued Over Training Data for Its AI and Face-Recognition Systems

Meta's AI training data practices could lead to significant financial penalties and force changes in how it uses user-generated content for AI development.

Meta faces a proposed class-action lawsuit alleging violations of Illinois and California privacy laws related to its AI and face-recognition systems. The suit claims Meta illegally extracted biometric information from user photos without consent for its NameTag feature, which was embedded in its AI companion app. Although the feature was not enabled, analysis indicated it was designed to create biometric signatures from faces captured by Meta glasses and compare them against a database on the user's phone, potentially updated by Meta. The complaint suggests this faceprint data may originate from Facebook and Instagram images. Additionally, the lawsuit targets Meta's image-generation systems, such as Emu, alleging that training data derived from Facebook and Instagram images unlawfully harvested biometric information. This action follows previous settlements and fines against Meta for mishandling biometric data, including a $650 million settlement in 2020 and a $1.4 billion agreement with Texas in 2024.

This lawsuit highlights the ongoing legal and ethical challenges of using vast datasets, including social media content, to train powerful AI models. The plaintiffs' claim that Meta harvested biometric information for both its generative AI and face-recognition systems suggests a broad concern about data privacy across Meta's AI ventures. The potential for millions of individuals to seek damages under stringent state privacy laws like Illinois' Biometric Information Privacy Act could impose substantial financial liabilities on Meta, influencing future data collection and usage policies across the industry.

The legal action against Meta's AI training data could set a precedent for how tech companies handle user-generated content when developing AI. If successful, such lawsuits might compel companies to seek more explicit consent for data usage in AI training, potentially slowing down development cycles or increasing compliance costs. This could also spur greater scrutiny of AI model training methodologies and data provenance, pushing the industry towards more transparent and privacy-preserving AI development practices, especially concerning biometric data.