Meta Platforms Inc. has discreetly embedded facial-recognition technology into the software powering its Ray-Ban and Oakley smart glasses, signaling a potential return to a controversial capability the company publicly claimed to have abandoned years ago. According to a technical analysis of Meta’s AI companion app, code for a feature internally designated as “NameTag” has been incrementally added through multiple software updates throughout 2024. When fully operational, the system is designed to identify individuals captured by the glasses’ integrated cameras and provide the wearer with real-time notifications identifying the person in their field of vision.
The discovery, first identified through a deep-dive analysis of the Meta AI app’s source code, reveals a significant discrepancy between the company’s internal technical developments and its public-facing communications. While Meta executives stated as recently as April 2024 that the company was merely “thinking through” the implications of facial recognition, core components of the recognition pipeline had already been distributed to millions of user devices as early as January. The Meta AI app, which is required to operate the smart glasses, has been downloaded more than 50 million times, effectively placing the building blocks of a mass-surveillance tool in the pockets of a global user base.
Technical Architecture of the NameTag System
The NameTag feature operates through a sophisticated three-stage AI pipeline that has already been deployed to user devices. Security researchers who independently verified the software’s structure found that Meta has distributed three distinct AI models to handle the biometric processing. The first model is tasked with detecting faces within the camera’s frame; the second model crops these faces to isolate them from the background; and the third model translates the visual data into a unique biometric signature, often referred to as a “faceprint.”
Once a face is converted into a biometric signature, the app is configured to check that signature against a local database stored on the user’s phone. This database is designed to receive updates from Meta’s servers, though it remains unclear exactly whose biometric data would be included in these updates or how the profiles are initially generated. Technical tests conducted by independent researchers, including the reverse-engineer known as Buchodi, demonstrated the system’s current efficacy. By manually adding a faceprint of the deceased philosopher Michel Foucault to the app’s internal gallery, researchers were able to trigger a “Person recognized” notification, confirming that the matching engine is functional despite being hidden from the standard user interface.
Further analysis of the code suggests that Meta is preparing to rebrand the feature for public consumption. In software versions released in May 2024, the "NameTag" moniker was replaced with "Connections," accompanied by marketing strings inviting users to "remember the people you met." Faces that the system does not immediately recognize are not discarded; instead, they are cropped, indexed, and stored in a folder marked “pending,” suggesting a system designed for retroactive identification or future enrollment.
A Reversal of the 2021 Privacy Pivot
The integration of NameTag represents a stark departure from Meta’s 2021 announcement, in which the company pledged to shut down its face-recognition system on Facebook. At that time, Meta stated it would delete more than a billion faceprints—the digital templates used to identify users in photos—citing "growing concerns about the place of facial recognition technology in society." That decision followed a decade of intense legal and regulatory pressure.
In 2020, Meta (then Facebook) agreed to a $650 million settlement in an Illinois class-action lawsuit involving the Biometric Information Privacy Act (BIPA), which remains one of the strictest privacy laws in the United States. In 2024, the company reached a further $1.4 billion settlement with the state of Texas over allegations that it had unlawfully collected biometric data without informed consent. Furthermore, in 2019, the company paid a record $5 billion fine to the Federal Trade Commission (FTC) to settle various privacy violations, including those related to facial recognition.
Despite these massive financial penalties and public commitments to privacy, internal documents suggest that the retreat from facial recognition was never intended to be permanent. Former Meta Reality Labs policy officials have indicated that there was persistent internal tension regarding when and how to reintroduce the technology. Documents leaked earlier this year indicated that Meta leadership sought a "dynamic political environment" to launch the feature on smart glasses, hoping that external political distractions would minimize the backlash from privacy advocates and regulators.
Chronology of Development and Public Statements
The timeline of Meta’s facial recognition development highlights a calculated rollout of code alongside a cautious public relations strategy:
- November 2021: Meta announces the "sunsetting" of its facial recognition system and the deletion of one billion faceprints.
- January 2024: Core components of the "NameTag" recognition pipeline begin appearing in updates to the Meta AI app.
- February 2024: Internal documents are leaked to the media, revealing Meta’s plans to debut facial recognition on smart glasses.
- April 2024: Over 70 privacy advocacy groups, including the ACLU and the Electronic Frontier Foundation (EFF), sign an open letter demanding Meta abandon the development of facial recognition for wearables.
- April 2024 (Response): A Meta spokesperson tells the media that "if we were to utilize face recognition, it wouldn’t be rolled out without first taking a very thoughtful approach."
- May 2024: The "NameTag" feature is rebranded as "Connections" within the app code, emphasizing social "remembering" over identification.
Privacy Implications and the Risk of "Distributed Surveillance"
The primary concern among privacy experts is the transformation of consumer hardware into a "distributed surveillance machine." Unlike stationary CCTV cameras, smart glasses are mobile, often indistinguishable from regular eyewear, and capable of capturing data in private or semi-private spaces where people have a high expectation of privacy.
Cooper Quintin, a senior technologist at the Electronic Frontier Foundation, warns that the technology could empower stalkers, domestic abusers, or unauthorized government agents to identify strangers in public without their knowledge or consent. Because the glasses do not have a prominent recording indicator that is visible from a distance, the "silent" nature of the identification process is particularly concerning.
There is also the psychological impact of "normalization." Legal scholars, such as Woodrow Hartzog of Boston University, argue that by embedding these features in popular consumer products, Meta is shifting societal norms. As biometric scanning becomes a routine part of social interaction, the public may lose its moral or ethical resistance to the technology. Hartzog notes that "consent" in these scenarios is often an illusion, as users may feel compelled to agree to biometric scanning to access necessary services or social benefits.
Assistive Technology vs. General Identification
Meta has partially defended its exploration of facial recognition by highlighting its potential as an assistive tool. For the blind and visually impaired community, the ability to identify people in a room is a significant accessibility breakthrough. A 2018 study conducted by researchers from Cornell Tech and Facebook found that blind participants viewed person-recognition as a critical daily task.
While existing assistive devices already allow users to "enroll" specific faces (such as family members) for identification, Meta’s system appears designed for a much broader scope. Critics argue that while the assistive use case is valid, it should not serve as a "Trojan horse" for a general-purpose identification system that could be used against the general public.
Official Response from Meta
Meta has maintained a stance of "exploration" rather than "deployment." Ryan Daniels, a spokesperson for Meta, stated that the presence of the code is merely evidence of the company’s ongoing research and development process. "Nothing has shipped to consumers and no final decision has been made on what to do here, if anything," Daniels said. He emphasized that if the company decides to move forward, it will do so with "full transparency" and a "thoughtful approach."
Significantly, Meta has clarified that it is not building a "central face database." The current code structure supports this claim to an extent, as the biometric matching is designed to occur on the user’s device. However, the fact that the device-side database is configured to receive updates from Meta’s servers suggests that the company still controls the source and flow of the biometric templates.
Broader Impact on the Wearable Industry
As Meta continues to lead the market in smart glasses through its partnership with EssilorLuxottica, its technical choices set the standard for the entire industry. The integration of NameTag suggests that the future of augmented reality (AR) and smart wearables will be inextricably linked to biometric data.
The legal and ethical questions remain unresolved. Will Meta require "opt-in" consent from every person captured by the glasses, or only from the wearer? How will the company navigate the patchwork of international biometric laws, such as the EU’s Artificial Intelligence Act, which places strict limits on remote biometric identification? As the code sits "nearly ready to go" on millions of devices, the transition from a private experiment to a public reality may only require a single software toggle, forever altering the landscape of public anonymity.
