Meta Sued Over Ray-Ban Smart Glasses Privacy
Meta is facing a class-action lawsuit over its AI-powered Ray-Ban smart glasses, alleging the company misled users about privacy. The suit claims footage was reviewed by human contractors for AI training, despite marketing that promised user control and privacy-first design. The case highlights the growing legal and reputational risks for companies deploying consumer-facing AI without transparent data-handling policies.
The class-action lawsuit, filed in the Northern District of California, names two plaintiffs: Gina Bartone of New Jersey and Mateo Canu of California. The suit alleges that Meta's marketing of the glasses as "designed for privacy, controlled by you" was intentionally deceptive. At the core of the issue is the use of human contractors in Kenya to review and label footage captured by the glasses to train Meta's AI. An investigation by Swedish newspapers revealed that these contractors viewed sensitive content, including nudity and private conversations, which sparked the lawsuit and regulatory interest from bodies like the U.K.'s Information Commissioner's Office. The AI in the Ray-Ban glasses is an extension of Meta's Llama models, specifically optimized for a hands-free, on-the-go experience and integrated with a "Hey Meta" voice assistant. This functionality relies on processing visual and audio data to identify objects and respond to user queries, and Meta's terms of service note that this data may be subject to human review to improve the AI. For developers, this case highlights the trade-offs between different AI training methodologies. While human-in-the-loop annotation can be effective for capturing nuance and mitigating bias, it poses significant privacy risks, especially with personal data from wearable devices. Alternative, more privacy-preserving techniques exist, such as federated learning, where the model is trained on-device without sending raw data to a central server. Another approach is the use of synthetic data, which can scale cost-effectively and avoid using real personal information, though it may lack the nuance of human-annotated data for certain complex tasks. In an April 2025 policy update, Meta made AI features on the glasses enabled by default and removed the option for users to prevent their voice recordings from being stored. While Meta states that media remains on the user's device unless they choose to share it, the lawsuit argues that the "choice" to interact with the AI assistant was not presented with a clear understanding that human review of sensitive moments was a possibility.
