Meta Tapped a Pentagon Supplier to Prototype Face Recognition for Its Glasses

Rank One Computing, which recently transitioned to a public company on the Nasdaq, derives approximately 80 percent of its revenue from government contracts. Its client list includes the U.S. Marshals Service and the Naval Criminal Investigative Service (NCIS). The firm has previously developed specialized long-range facial recognition for the U.S. Special Operations Command, claiming the ability to identify individuals from distances as great as one kilometer. The integration of such technology into a consumer device like smart glasses raises significant questions regarding privacy, consent, and the future of public anonymity.

The Scope of the Rank One Partnership

The licensing agreement between Meta and Rank One Computing provides a detailed look at the technical ambitions Meta has for its hardware. The license authorizes the use of Rank One’s facial recognition algorithms alongside "liveness detection" technology. Liveness detection is a security feature designed to distinguish between a living human face and a high-resolution photograph or a 3D mask, ensuring the system cannot be easily spoofed.

According to the technical documentation, the system was configured to support up to 10 million facial templates. These templates are digital representations of facial features used to compare a captured image against a database of known identities. While Meta’s smart glasses are currently marketed as tools for hands-free photography, video streaming, and AI-assisted tasks, the inclusion of Rank One’s code suggests that Meta has actively experimented with turning these glasses into real-time identification tools.

Code analysis of the Meta AI app revealed that although the Rank One integration was dormant in the version shipped to millions of consumers, the underlying routines required to initialize the software remained present until very recently. This discovery follows reports of an unreleased internal Meta project dubbed "NameTag," which was designed to allow users to identify people in their field of vision by matching their faces against Meta’s vast social media databases.

A History of Surveillance and Government Ties

Rank One Computing was founded in 2015 by a group of engineers with deep roots in the defense and intelligence sectors. The founding team previously worked at Noblis, a nonprofit research institute where they evaluated biometric algorithms for U.S. intelligence agencies. Since its inception, Rank One has positioned itself as a premier provider of "mission-critical" biometric solutions.

The company’s leadership reinforces its status as a bridge between the private sector and the national security establishment. Chief Executive Officer B. Scott Swann is a former official at the FBI, where he managed the division responsible for the bureau’s massive biometric databases. The company’s board of directors is equally prestigious within the intelligence community, featuring a former CIA deputy director for science and technology, a former head of the FBI’s science and technology branch, and a former Pentagon official who oversaw a multi-billion-dollar office for special capabilities.

Rank One’s technology is already deployed in various sensitive environments:

• U.S. Marshals Service: Since 2021, the agency has used Rank One’s biometric identification kits to confirm the identities of prisoners during transport, bypassing the need for traditional fingerprinting.
• Naval Criminal Investigative Service: The Navy’s primary law enforcement arm utilizes "ROC Watch," a video analytics tool designed for real-time monitoring.
• Public Education: In West Virginia, Rank One’s software has been implemented in dozens of schools to screen visitors against state sex-offender registries.
• Law Enforcement Networks: The company’s algorithms are embedded in third-party tools like LexisNexis’s Lumen platform and DataWorks Plus, which allow local police officers to run facial searches against state, regional, and FBI databases.

Chronology of Integration and Deletion

The timeline of Meta’s engagement with Rank One Computing suggests a period of intensive testing followed by a rapid retreat once the project’s existence became public knowledge.

• 2015–2020: Rank One Computing establishes its presence as a key vendor for U.S. law enforcement and military agencies.
• 2021: Meta (then Facebook) announces it will shut down its "Face Recognition" system on its social media platform, citing "growing societal concerns."
• 2023–Early 2024: Meta begins integrating Rank One’s SDK (Software Development Kit) and facial recognition routines into the Meta AI app, the companion software for its Ray-Ban smart glasses.
• February 2024: Rank One Computing goes public on the Nasdaq.
• June 4, 2024: Reports emerge detailing "NameTag," an unreleased facial recognition system built into the Meta AI app, which has been downloaded over 50 million times.
• June 5, 2024: Within 24 hours of the public revelation, Meta deletes the Rank One license routines and all facial recognition code from the Meta AI app.

Meta has remained largely silent regarding the specific nature of its relationship with Rank One. The company declined to clarify why it chose a military-grade vendor over developing its own internal solutions or when the licensing agreement was first signed. Similarly, Rank One Computing has declined to comment on its business dealings with the social media giant.

Technical Performance and Demographic Disparities

A critical concern regarding the proliferation of facial recognition is the accuracy and bias of the underlying algorithms. While Rank One markets its software as highly accurate, independent testing by the National Institute of Standards and Technology (NIST) has revealed significant performance disparities.

In NIST’s Face Recognition Vendor Test (FRVT), Rank One’s algorithms demonstrated varying error rates depending on the demographic group being analyzed. The software produced false matches at notably different rates based on sex and country of birth. Specifically, error rates were found to be lowest for individuals born in Eastern Europe—a group often used by NIST as a proxy for certain racial characteristics—while error rates were consistently higher for women than for men.

The "liveness detection" feature included in Meta’s license is intended to mitigate some security risks, but it does not address the fundamental issue of demographic bias. If deployed in a consumer device, such inaccuracies could lead to "false positives," where a user incorrectly identifies a stranger, or "false negatives," where the system fails to function for certain users based on their physical appearance.

Broader Implications for Privacy and Regulation

The revelation of the Meta-Rank One partnership underscores the lack of a cohesive regulatory framework for biometrics in the United States. While several states, such as Illinois and Texas, have implemented stringent biometric privacy laws, there is no federal statute governing how private companies or law enforcement agencies can use facial recognition.

Eric Null, director of the Privacy and Data Project at the Center for Democracy and Technology, noted that consumer-facing companies are increasingly "craving access" to high-powered surveillance technology. "Without proper checks, the risks of this tech becoming a common consumer product are significant and largely unbounded," Null stated.

The transition of military technology into the consumer sector is a well-documented phenomenon. Joseph Jerome, a former policy official at Meta Reality Labs, observed that the history of the internet itself is a story of military innovation becoming a public utility. However, the move toward real-time, wearable facial recognition represents a new frontier. Unlike a smartphone, which must be held up to take a photo, smart glasses allow for "ambient" data collection, where the wearer can record or identify people without their knowledge or consent.

Analysis: The Convergence of Consumer and Surveillance Tech

The integration of Rank One’s software into Meta’s ecosystem suggests that the "creeper" factor—a term used by critics to describe the social discomfort caused by camera-equipped glasses—is only the beginning. If Meta or its competitors successfully normalize facial recognition in wearables, the very concept of "walking down the street anonymously" could disappear.

From a business perspective, Meta’s interest in this technology is logical. Providing users with the ability to instantly "know" who they are looking at—by pulling up a LinkedIn profile or a Facebook page—would be a "killer app" for augmented reality. However, the ethical cost is immense. By sourcing this capability from a company like Rank One, Meta is effectively leveraging tools designed for the battlefield and the precinct to enhance a consumer fashion accessory.

The rapid deletion of the code on June 5 suggests that Meta is aware of the potential for public backlash. Nevertheless, the fact that the code was present in a "shipped" version of the app indicates that the company was closer to a launch than its public statements on privacy might suggest. As smart glasses continue to evolve from niche gadgets into mainstream devices, the pressure to include advanced biometric features will likely grow, further blurring the line between personal convenience and state-level surveillance.