The U.S. Government Adopted AI Smart Glasses Before Any Law Did

Meta’s AI smart glasses have been carrying a hidden payload. A Wired investigation published June 4 found dormant facial recognition code, internally called NameTag, inside the companion app on more than 50 million phones. Key components arrived as early as January, months before any public disclosure, and security researchers who reviewed the code told Wired the feature looked nearly ready to activate. The U.S. has no federal law written specifically for AI wearables.

That gap has been filled by something Congress never voted on. Immigration and Customs Enforcement agents have been wearing the same consumer glasses in field operations across at least six states since the start of President Donald Trump’s second term in January 2025. The Department of Homeland Security has asked Congress for $7.5 million to build its own version, with real-time biometric identification built in.

Identifying Strangers Before the Next Stop

In October 2024, Harvard students Caine Ardayfio and a classmate walked through public spaces wearing Meta’s Ray-Ban smart glasses. They’d built a system called I-XRAY: the glasses’ live feed ran through facial recognition software, which pulled names, addresses, and phone numbers from public databases for strangers who had no idea they were being scanned.

“We’re able to identify dozens of people using our glasses, including Harvard students, without them ever knowing,” Ardayfio said.

I-XRAY was a proof of concept. Its authors published the project as a warning. The hardware it ran on is already in millions of hands: Meta has sold more than 8 million pairs of its AI glasses since 2024, and the market keeps expanding despite growing public unease. Jitesh Ubrani, a research manager at International Data Corporation (IDC, a technology market intelligence firm), told International Business Times that restaurants and workplaces have started banning the devices. Sales keep climbing regardless. “The hesitation is also growing, but not enough to offset sales,” Ubrani said.

The glasses look like ordinary sunglasses. They record video, take photos, livestream footage, and run AI queries on whatever the wearer is looking at in real time. The current Ray-Ban model carries Meta’s Live AI feature, which can describe a surrounding environment and respond to questions about it while the wearer walks. The recording indicator light the hardware ships with can be disabled by modders, removing the only signal bystanders would have that a camera is running.

NameTag’s Six-Month Secret

The investigation Wired published June 4 found that Meta had been distributing dormant facial recognition code inside its Meta AI companion app since January. The system, internally called NameTag, was built to recognize faces through the glasses’ camera and notify the wearer when a previously saved person appears nearby. Wired shared its findings with security researchers, including Cooper Quintin, a senior public interest technologist at the Electronic Frontier Foundation’s (EFF) Threat Lab, who said the feature appeared nearly ready to launch.

• 50 million+ phones carry dormant facial recognition code through the Meta AI app, distributed since January 2026
• 3 AI models in the NameTag pipeline handle face detection, cropping, and biometric encoding separately
• 0 public disclosures: the code reached millions of devices before Wired’s investigation appeared

The on-device architecture is the engineering choice that makes the system both fast and legally complicated. By matching faceprints on the phone rather than streaming them to the cloud, Meta can argue it isn’t maintaining a central facial recognition database. Wired’s review found the code also draws from Meta’s servers, making that distinction contested. A second researcher tested the system by adding a faceprint for a public figure and successfully triggered a recognition notification.

Faces the system recognizes trigger a notification to the wearer. Faces it doesn’t recognize are reportedly cropped, indexed, and saved to a background folder on the phone. Every bystander the wearer passes becomes an entry in that folder.

Meta spokesperson Ryan Daniels told multiple outlets that “nothing has shipped to consumers and no final decision has been made on what to do here, if anything.” The company said it would be transparent about any launch and isn’t building a central face database. An internal memo reviewed by reporters at Engadget noted that Meta saw a “dynamic political environment” as a strategic window for launch, because civil society groups opposing it “would have their resources focused on other concerns.” Working code distributed months earlier is difficult to square with a decision the company described as undecided.

The Forty-Year Gap

The Electronic Communications Privacy Act (ECPA), the main federal statute governing digital surveillance in the United States, was signed into law in 1986. Brookings, the Washington policy research institution, noted this year that Congress had expected the statute to last roughly a decade or two before needing an update. It has now reached its 40th birthday, untouched in many material ways in the interim. The law was written for a world of dial-up modems and local servers, with nothing to say about a device that harvests biometric data from strangers’ faces on a public sidewalk.

The result is a fragmented landscape that the Center for Democracy and Technology has called on Congress to address. Illinois and Texas have biometric data laws touching facial recognition. Nevada and New York have consent statutes. Colorado passed the first state-level AI privacy law. None were drafted with wearable AI cameras in mind, enforcement varies across state lines, and no federal bill specifically targeting AI wearables has become law.

Three Senate Democrats have pressed Meta directly on the NameTag plans. Senators Ed Markey, Jeff Merkley, and Ron Wyden sent Meta an open letter warning the company about the consequences of its facial recognition roadmap:

This frictionless identification and constant monitoring risks normalizing mass surveillance at a moment when the federal government is using similar tools to intimidate protesters and chill speech.

The letter was addressed to Meta. By the time it was published, federal agents had already been wearing the same consumer glasses in field operations for months.

Agents in Six States, No Authorization

Investigations by The Independent and 404 Media found Immigration and Customs Enforcement (ICE) and Customs and Border Protection (CBP) agents wearing Meta’s Ray-Ban smart glasses during field operations across at least six states since Trump’s second term began. One Border Patrol agent was photographed at a raid near a Home Depot in Los Angeles in June 2025, recording indicator light clearly on. In Evanston, Illinois, agents were observed filming protesters with the same glasses, indicator lights active.

DHS regulations specifically prohibit using personal recording devices to capture individuals engaged in First Amendment-protected activities without reasonable suspicion. Official body cameras operate under defined retention schedules and are subject to public records requests. Consumer glasses connected to a personal Meta account route footage to Meta’s servers under consumer terms of service, with no equivalent retention requirement or audit trail. Accessing what an agent’s glasses captured would require separate legal process, not a standard agency records request.

The glasses can do more than record passively. They livestream directly to the internet and use AI to analyze the wearer’s field of vision in real time. Pairing that feed with external facial recognition software, as researchers demonstrated in 2024, would require no additional hardware beyond the glasses themselves.

Incidents documented in the press show how the use has unfolded on the ground. In January 2026, an ICE agent in Maine allegedly photographed the license plate of a woman who had been filming arrest operations earlier that day; she subsequently joined a class action lawsuit challenging the Trump administration’s conduct toward people who observe or film immigration operations. An agent in a separate incident told a bystander: “We have a nice little database and now you’re considered a domestic terrorist.” DHS denied maintaining any database cataloging people who film or protest its operations.

DHS’s $7.5 Million Plan

The unauthorized consumer use was the preview. The Department of Homeland Security’s fiscal year 2027 budget justification for its Science and Technology Directorate, filed in April, requested $7.5 million to develop “critical technologies, analytic tools, and data systems” for detention and removal operations. Among the line items: “operational prototypes of smart glasses” designed to give agents “real-time access to information and biometric identification capabilities in the field,” per the budget document reviewed by journalist Ken Klippenstein.

The request reached Capitol Hill without advance notice to most members. “I’d have to look at it,” said Senator Thom Tillis, a Republican on the Senate homeland security panel, after hearing the details. Representative Carlos Gimenez of Florida, who sits on the House Homeland Security Committee, replied when asked whether recording civilians on the street concerned him: “There is no expectation of privacy when you’re in the street.”

DHS told reporters in April that no federal funds had been committed for any smart glasses program. Its Science and Technology Directorate is “constantly assessing” field needs, a spokesperson said, and any technology adopted would remain “within the full scope of the law.” ICE set a target deployment date of September 2027 for the technology, per Klippenstein’s reporting.

The agency’s own trajectory makes the direction plain. DHS’s Mobile Fortify app, which enables facial recognition through a phone camera, had been used more than 100,000 times since its June 2025 launch, per a January 2026 lawsuit filed by the State of Illinois and the City of Chicago against DHS and then-Secretary Kristi Noem. Mobile Fortify queries against federal databases including DHS’s IDENT biometric system, which holds more than 270 million records. DHS’s own 2025 AI Use Case Inventory classified Mobile Fortify as “high-risk” and noted it was deployed without the legally required Privacy Impact Assessments. Smart glasses would place that same query capability directly in an agent’s line of sight, hands-free. A DHS attorney, speaking anonymously to Klippenstein, warned the planned glasses would affect “all Americans, particularly protesters.”

Counter-Surveillance on a Shoestring

On March 5, the Clarkson Law Firm, a California-based public interest firm, filed a class action lawsuit against Meta in the U.S. District Court for the Northern District of California. Plaintiffs Gina Bartone and Mateo Canu alleged Meta had marketed the glasses as “designed for privacy” while transmitting footage to offshore contractors in Nairobi, Kenya for AI training without disclosing the practice. An investigation by Swedish newspapers Svenska Dagbladet and Göteborgs-Posten found those contractors had viewed intimate footage from users who appeared not to know they were recording.

Civil society responses have been building on several tracks since early in the year:

• The American Civil Liberties Union (ACLU) and the Electronic Privacy Information Center (EPIC, a public interest research center) urged the Federal Trade Commission (FTC) to halt Meta’s facial recognition plans, calling them a “grave risk to privacy, safety, and civil liberties.”
• Texas Attorney General Ken Paxton pledged to “relentlessly stand up to any company that threatens the privacy and safety of Texans.”
• Senators Markey, Merkley, and Wyden introduced the ICE Out of Our Faces Act in February, which would ban ICE and CBP from using facial recognition entirely.
• No federal legislation specifically targeting AI wearables has cleared either chamber of Congress.

Yves Jeanrenaud, a sociology professor at Osnabrueck University of Applied Science in Germany, decided not to wait. He built a free Android app called Nearby Glasses in roughly eight hours. It scans for Bluetooth Low Energy (BLE) signals emitted by smart glasses from Meta, Snap, and Luxottica Group and alerts nearby users when one is detected. The source code is published on GitHub and open to contributions from any developer.

The app’s limits reflect the scale of the underlying problem. It runs only on Android, fires false alerts on VR headsets sharing the same manufacturer Bluetooth signals, and is useless when a device has Bluetooth disabled. “Of course, it’s a technical solution to a social problem,” Jeanrenaud told TechCrunch. “I do not want to promote techsolutionism, nor do I want people to feel falsely secure.” In 2024, Apple and Google built cross-platform Bluetooth detection for AirTag location trackers. No equivalent industry collaboration exists for smart glasses, leaving detection entirely to independent developers working without a budget.

The DHS request for its own biometric glasses arrived on Capitol Hill in April. Agents had been using Meta’s consumer version in field operations for more than a year by then, without department authorization and without a dedicated federal law covering anything they captured.