Meta's Ray-Ban Smart Glasses Have a Privacy Problem

From across a room, the Ray-Ban Meta AI smart glasses look like ordinary frames. That is the entire point. They cost between $300 and $800, translate languages on command, capture sunsets without the awkwardness of holding up a phone, and have sold by the millions. They also route a portion of what they record through a global pipeline of human reviewers, some of whom have spent the last year watching strangers undress in their own bedrooms.

In February, the Swedish newspapers Svenska Dagbladet and Göteborgs-Posten published an investigation that should have rattled every person who has ever said “Hey Meta” to a pair of the company’s smart glasses. Reporters traveled to Kenya and interviewed more than thirty workers at Sama, a data annotation subcontractor hired by Meta to train the AI built into its Ray-Ban and Oakley smart glasses. What those workers described was a steady stream of intimate footage from Western homes, captured by users who, in many cases, did not appear to know they were filming at all.

“We see everything – from living rooms to naked bodies,” one worker told the reporters. “Meta has that type of content in its databases. People can record themselves in the wrong way and not even know what they are recording. They are real people like you and me.”

The workers described people walking out of bathrooms naked. Couples in bed. Users watching pornography while wearing the glasses. Bank cards inadvertently filmed. Text conversations covering crimes, personal secrets, and explicit sexual commentary. They described offices monitored by security cameras, personal phones banned at the door, and non-disclosure agreements that turned questions into firing offenses. “You understand that it is someone’s private life you are looking at,” one worker said, “but at the same time you are just expected to carry out the work. You are not supposed to question it. If you start asking questions, you are gone.”

This is not a leak. This is policy.

Meta sells the Ray-Ban glasses as a sleek everyday assistant. Voice-activated. Always ready. Privacy-first. CEO Mark Zuckerberg demonstrated them onstage in September 2025 at Meta Connect, walking through corridors of his Menlo Park headquarters as the audience watched the world through his eyes. The pitch was seductive: an all-in-one assistant that could translate languages, identify landmarks, and capture moments hands-free. Meta sold seven million pairs in 2025, a more than three-fold jump from the previous two years combined.

What Zuckerberg did not say onstage is what Meta’s terms of service say in writing. Buried in the AI Terms of Service is language authorizing Meta to review user interactions, including the content of conversations and messages, and stating that this review “can be automated or manual (human).” When the Swedish journalists tested the glasses themselves, they found they would not function without an active internet connection, and that data was automatically routed through Meta’s servers regardless of which privacy options the user selected during setup. When the reporters visited ten eyewear retailers in Stockholm and Gothenburg, sales staff gave contradictory answers about where customer data goes. Several insisted nothing was shared with Meta at all.

Several were wrong.

The disconnect between what consumers think they are buying and what they are actually agreeing to is the foundation of the entire arrangement. Most people do not read terms of service. Almost no one reads them carefully. And Meta has structured its disclosures so that the existence of human reviewers is technically present in the documentation, while the practical reality of that review, the volume of footage involved, the geography of the workforce, and the kinds of intimate moments being scrutinized, sits well outside what any reasonable consumer could be expected to anticipate.

The Swedish investigation triggered a regulatory response. The UK’s Information Commissioner’s Office contacted Meta about the “concerning” reports. Kenya’s Office of the Data Protection Commissioner announced an investigation of its own. A consumer protection lawsuit was filed in March.

Meta’s response came two months later, and it was not the response anyone watching was expecting. Rather than reform its data review pipeline, audit the Sama operation, or publicly clarify what its workers see and how they handle it, Meta terminated its entire contract with Sama. The company told the BBC it had ended the relationship because Sama “didn’t meet our standards.”

Naftali Wambalo of the Africa Tech Workers Movement, who has spoken with workers involved in the Meta glasses project, offered a different reading. “What I think are the standards they are talking about here are standards of secrecy,” Wambalo told the BBC. The implication: Meta did not punish a contractor for failing to protect users. It punished a contractor whose workers talked to reporters.

The footage did not stop being collected. The annotation work did not stop being necessary. Sama announced 1,108 layoffs at its Nairobi facility, with no public word from Meta on where the work is going.

The pattern showed up again a few weeks earlier, in a Los Angeles courtroom. During the landmark trial over social media’s impact on children, members of Zuckerberg’s team walked into the courtroom wearing Ray-Ban Meta AI glasses. Judge Carolyn Kuhl ordered them removed and warned the team that anyone who had recorded anything needed to dispose of it or face contempt.

If the existing privacy concerns sound like the worst of it, they are not. In February of this year, The New York Times reported that Meta has been developing a face-recognition feature for its smart glasses, known internally as “Name Tag.” The feature would let a wearer look at a stranger and pull up information about them through Meta AI. Engineers have reportedly been weighing two versions: one limited to people the wearer is already connected to on a Meta platform, and a broader version that could identify anyone with a public Meta account.

In May 2025, according to internal documents obtained by the Times, a Reality Labs memo proposed launching the feature “during a dynamic political environment where many civil society groups that we would expect to attack us would have their resources focused on other concerns.”

Read that sentence again. Meta did not write it accidentally. A senior team at one of the largest companies in the world looked at rising authoritarianism and the dismantling of regulatory capacity in the United States and identified it as a market opportunity. They proposed timing the rollout to coincide with civil society being too overwhelmed to push back.

In April 2026, a coalition of more than seventy organizations, including the ACLU, the Electronic Privacy Information Center, Fight for the Future, and the Leadership Conference on Civil and Human Rights, wrote to Zuckerberg demanding he kill the feature before launch. The coalition argued that face recognition embedded in inconspicuous consumer eyewear “cannot be resolved through product design changes, opt-out mechanisms, or incremental safeguards.” Bystanders in public have no meaningful way to consent to being identified by a stranger’s glasses. Senators Ron Wyden, Ed Markey, and Jeff Merkley sent their own open letter raising the same alarm, warning that the technology could “capture images of thousands of people without their knowledge or consent and then instantly link those faces to names, workplaces, or personal profiles, creating serious risks of stalking, harassment, and targeted intimidation.”

A Meta spokesperson told Wired, “Our competitors offer this type of facial recognition product, we do not. If we were to release such a feature, we would take a very thoughtful approach before rolling anything out.” In other words, the company is not denying the feature exists in development. It is denying that it has been released yet.

While Meta and its critics argue about a feature that has not yet shipped, a separate problem is already operating at scale. The Ray-Ban glasses ship with a small LED light in the upper corner of the frame, intended to alert bystanders when recording is active. That light has become a target.

A YouTuber with 23,000 subscribers has demonstrated a hack for blocking the LED. A TikTok account in Southern California advertises a “stealth mode” service that physically removes the LED from a customer’s glasses for $120. And a wave of social media accounts now exists for the express purpose of recording strangers without their knowledge. Wired documented a growing subculture of self-styled pickup artists, including influencers with millions of combined Instagram followers, who use the glasses to film their attempts to chat up women in nightclubs, on beaches, and on city streets. In many of the videos, the women appear visibly uncomfortable, and they are clearly unaware they are being filmed.

It is not difficult to see where this leads. A man approaches a woman on the street, films her without consent, and uploads her face to an audience of millions. Currently, that man has to do the legwork himself. If Name Tag ships, he gets her name, workplace, and social media profiles delivered to him within seconds, all while the woman is still trying to figure out why a stranger is talking to her.

The tools for stalking already exist. Smart glasses are about to make them frictionless.

It is tempting to read all of this as a privacy story. It is, but the framing undersells it. The deeper question is whether anonymous public life is something we still get to have.

For most of human history, the answer was yes by default. You could walk down the street and be a stranger. You could attend a protest, a religious service, a recovery meeting, or a doctor’s appointment, and not be cataloged on the way in. The anonymity was not perfect but it was real enough to live by. What smart glasses propose is the elimination of that condition for everyone, decided unilaterally by a handful of tech companies, executed through a consumer product that millions of people are buying without quite understanding what comes with it.

Meta is not the only company in this space. Google has partnered with Warby Parker to develop AI-powered smart glasses of its own. Apple is rumored to be developing a competing product. Snap already sells Spectacles. The market is consolidating around a single proposition: the camera should be on your face, the AI should be in the cloud, and your data should belong to whoever is selling you the hardware.

We have seen this pattern before. A new technology arrives. Early concerns are dismissed as overreaction. The technology becomes ubiquitous before regulation catches up. By the time a meaningful response is possible, the social norms have already shifted, and the people who tried to warn anyone are scolded for being slow to embrace the future. Smartphones did this. Social media did this. Facial recognition did this on a smaller scale, until Meta itself paid roughly $2 billion to settle biometric privacy lawsuits in Illinois and Texas over its old Facebook photo-tagging system.

Meta paid that money and shut down photo-tagging in 2021. The company’s own statement at the time framed the decision as “a company-wide move away from this kind of broad identification,” pending clearer regulatory guidance. Five years later, that same company is preparing to relaunch the same capability in a more invasive form, on a device people wear on their faces in public. The internal memo describing that plan does not pretend otherwise. It treats civil society pushback as a scheduling problem.

A Kenyan annotator told the Swedish journalists what is probably the most clarifying line in the entire investigation. “You think that if they knew about the extent of the data collection, no one would dare to use the glasses.”

The business model depends on you not knowing. The marketing depends on you not asking. The terms of service are written so that “human review” is technically disclosed and practically invisible. The Sama contract was terminated not because the workers were wrong, but because they spoke out. The Name Tag memo treats the erosion of public anonymity as a launch window. And the LED light that was supposed to protect bystanders has been turned into a $120 aftermarket service for the people who want to defeat it.

None of this is hypothetical. All of it is happening now, in 2026, in homes and on sidewalks across the United States. The question is no longer whether smart glasses will reshape public life. They already are. The question is whether the rest of us get any say in the terms.

WHAT YOU CAN DO NOW

The privacy controls inside the Meta View app are useful, but they are not the answer. Meta will not change course because individual users opted out of cloud media sharing. The company will change course when the legal and regulatory environment makes the current arrangement too costly to maintain. That requires pressure on legislators and regulators, not just on settings menus.

Three asks, in order of leverage:

1. Push back on Name Tag before it ships. Senators Ron Wyden, Ed Markey, and Jeff Merkley have already written to Mark Zuckerberg demanding answers about Meta’s facial recognition plans. EPIC has filed formal complaints with the FTC and state enforcers asking them to block the rollout. Contact your senators and ask them to add their names to the Wyden-Markey-Merkley effort. Contact the FTC and tell them you want Meta’s biometric plans investigated before launch, not after.
2. Support a federal biometric privacy law. Only Illinois has a biometric privacy statute strong enough to let individuals sue, and Meta has paid roughly $2 billion to settle violations under that law and a similar one in Texas. Most states have far weaker protections or none at all. Federal versions, including one introduced by Senator Merkley in 2020, have repeatedly stalled in Congress. Contact your House representative and ask where they stand on biometric privacy legislation, and whether they will co-sponsor a federal version.
3. Demand enforceable disclosure standards for wearable recording devices. The LED light on Meta’s glasses is voluntary, defeatable, and governed by Meta’s own terms of service rather than law. Bystanders have no enforceable right to know they are being recorded. State legislators are where this fight is most winnable in the short term. Find out who chairs your state’s consumer protection or technology committee, and write to them.

And while you are doing all of that: open the Meta AI mobile app, go to Device Settings, then Glasses Privacy, and turn off Cloud Media. It will not save the world. It will withdraw your small piece of consent from a system that depends on consent at scale.

Privacy. Accountability. Clarity. Subscribe to get every piece in your inbox.