Meta's smart glasses have evolved from a consumer gadget into a critical data-collection tool for building artificial intelligence systems that can understand human behavior in real-world contexts. The Ray-Ban smart glasses, made in partnership with EssilorLuxottica, capture egocentric video showing exactly what wearers see, and Meta uses this footage to train AI models designed to power personalized AI assistants. However, the strategy has sparked significant backlash over privacy and consent, with critics dubbing the devices "pervert glasses" due to reports of nonconsensual recording.

How Did Meta's Strategy Shift From the Metaverse to AI?

Meta's pivot toward AI infrastructure began over a decade ago. In September 2013, Facebook announced its first artificial intelligence research lab focused on deep learning. That December, the company hired Yann LeCun, a deep learning expert from New York University's Courant Institute of Mathematical Sciences, to lead the effort. By 2019, Meta integrated the technical infrastructure underlying WhatsApp, Instagram, and Facebook Messenger, creating unified data flows across its entire platform ecosystem.

The metaverse strategy, which the company heavily promoted starting in 2021, proved unsuccessful by 2022. As ChatGPT and other large language models brought generative AI into the mainstream, Meta shuttered Horizon Worlds and redirected resources toward AI development. The company introduced its LLaMA model family in 2023, launched the Meta AI assistant in 2025, and released Muse in 2026, each step designed to draw on data generated across Meta's ecosystem.

What Data Are Smart Glasses Actually Collecting?

Ray-Ban smart glasses serve as primary data collection devices, capturing continuous first-person video streams. Meta uses this egocentric data to train what Zuckerberg calls "low-friction contextualized AI" interfaces, meaning AI systems that understand the specific context of what a user is seeing and doing.

The data collection extends beyond cameras. Meta also uses EMG neural bands, wearable devices that measure electrical signals from user muscles to infer intent and emotion without explicit user communication. This biosignal data feeds into AI training processes alongside video and behavioral information from Meta's social platforms.

The scope of data collection is broader than active users might realize. Even people who do not use Meta products can become data points when others photograph, record, or mention them through Meta systems. Privacy laws place limits on what Meta can collect, and users can opt out in some contexts, but vast quantities of data remain accessible.

Ways Meta Is Addressing Privacy and Safety Concerns

• LED Indicator Light: The Ray-Ban smart glasses feature an LED light that activates whenever someone captures a photo or video for their gallery, providing visible notification to people nearby that recording is occurring.
• Tamper Detection Technology: The glasses include tamper detection systems, and Meta is updating them to disable the camera if the LED is physically tampered with or destroyed, a safety feature the company says no other camera manufacturer has implemented.
• Safety Guidelines: Meta's official safety guidelines explicitly state that smart glasses should not be used for harassment, invading privacy, or recording sensitive information without consent.

"We designed our AI glasses with privacy and safety features built in from the start, informed by engagement with safety experts, including women's safety organizations. Our glasses have always had an LED light that activates whenever someone captures a photo or video for their gallery. They also feature tamper detection technology, and we are updating the glasses to disable the camera if they detect the LED was physically tampered with or destroyed. No other kind of camera has done this and we're proud to lead the industry forward," Meta stated in response to safety concerns.

Meta, company statement

Why Are Privacy Advocates Concerned?

Despite Meta's safety features, many women have reported on social media that they have been filmed without consent by men wearing the glasses. The devices have earned the colloquial nickname "pervert glasses" among internet critics due to reports of nonconsensual recording and sharing of images and videos on social media.

The deeper concern centers on how Meta intends to use the data generated through these devices. Zuckerberg's 2026 manifesto describes a world where personal AI agents will do users' bidding, but building those systems requires enormous amounts of data about human behavior. Critics argue that even seemingly incidental information captured through smart glasses can be swept into Meta's AI training systems, creating what one analyst described as an "algorithmic data farm".

The challenge of maintaining privacy in a world of ubiquitous, networked cameras may not be scalable. People who do not actively use Meta products cannot fully opt out from nonconsensual data capture if someone else records them. This means ordinary people are effectively helping finance Meta's AI ambitions through the forfeiture of privacy, whether they consent or not.

What Is Zuckerberg's Long-Term Vision?

Zuckerberg's goal is to create what he calls "low-friction contextualized AI," built around technologies including EMG neural bands and egocentric data from smart glasses. This vision ultimately relies on vast amounts of human behavioral data. The pieces of this vision have been assembled over more than a decade, from the 2013 founding of Facebook's AI research lab through the 2019 infrastructure integration to the 2021 metaverse announcement and the subsequent pivot to AI.

In essence, Meta's vision has not abandoned the metaverse so much as relocated it. Rather than building a separate virtual world users choose to enter, Meta is making the physical world interactive through AI systems trained on real-world behavioral data. The interface is increasingly the world around us, captured through smart glasses and other hardware, rather than a virtual environment.

This strategy creates a fundamental asymmetry. Meta benefits at multiple points in the data cycle, while users help finance the company's AI ambitions through the forfeiture of privacy. Even those who do not actively use Meta products become "algorithm chow," feeding the models intended to realize Zuckerberg's vision.