Meta Smart Glasses Face Recognition Code: Lawsuit and Leaked Memo Explained

Meta's Ray-Ban smart glasses are facing a federal class-action lawsuit alleging the company sold the product on explicit privacy promises while routing footage to human reviewers overseas through a pipeline buyers could not opt out of, per MediaLaws this month. Separately, a leaked internal memo reportedly describes Meta's interest in adding facial recognition code to the glasses a feature that would let wearers identify strangers in real time. The two stories are connected. Understanding why requires looking at what the product already does, and what adding identification capability would actually change.

A note on the record: the lawsuit draws its core evidence from a February 2026 investigation by Swedish newspapers Svenska Dagbladet and Göteborgs-Posten, the sole journalistic source cited in the complaint, per MediaLaws. The New York Times reported on the facial-recognition plans, based on an internal Meta document; the details below come from a Journal of High Technology Law summary of that reporting and from EFF analysis. Meta has not filed a formal defense. Facial-recognition deployment is not confirmed.

Two consent failures baked into the current product

The suit Bartone and Canu v. Meta, filed in the Northern District of California is built on consumer-protection claims, not privacy torts. False advertising, no meaningful opt-out, inadequate disclosure. That framing is specific and deliberate.

The complaint does not argue that footage review is inherently illegal. It argues that Meta's marketing set an expectation the product didn't honor. Buyers were told "Designed for privacy, controlled by you" and "You're in control of your data and content." What they were not clearly told, the suit alleges, is that using the glasses' AI features meant footage entered a review pipeline staffed by outside contractors, with no way to opt out and still retain the product's functionality. Decline the data arrangement and a buyer is left with what the complaint describes as "a $299 to $799 pair of sunglass frames, with no AI or smart functionality whatsoever," per MediaLaws.

Meta's position is that human review was disclosed in its supplemental privacy policy and, in the U.K., in AI terms of service stating that interactions "may be automated or manual (human)," per TechCrunch in March. The complaint's counterargument is that prominent product-page claims of thorough user control overwhelmed any fine-print disclosure. That is the question now before a federal judge: whether a company's loudest marketing can be contradicted by policy language most buyers never read.

The second failure involves people who never bought anything. The glasses carry a near-invisible camera and require only a casual touch to record, per BBC in May. No consent mechanism was designed for bystanders, because none exists for them by definition. The Swedish investigation found annotation workers were routinely shown footage of people undressing, using bathrooms, and handling financial documents the predictable output of a wearable camera left running in domestic settings, per MediaLaws. One account in the complaint describes a woman undressing while her partner's glasses sat on a bedside table. The automated face-blurring system meant to protect those bystanders reportedly failed frequently, leaving faces visible in footage sent to contractors, per MediaLaws.

More than seven million pairs sold globally in 2025 alone, per TechCrunch. Meta holds over 80% of all AI smart-glasses sales, per BBC. The consent architecture this company builds or fails to build will be the template every competitor designs around.

Bystander harm is already documented, no face recognition required

The bystander problem is not theoretical. Women have been filmed at beaches, outside shops, and on sidewalks by men wearing the glasses often while responding to conversation and discover the footage only after it circulates online alongside harassment, per BBC. Retail workers have been filmed reacting to candles sprayed with bad odors; drive-thru staff have been recorded during staged food thefts. The footage gets posted for engagement. None of it required identification capability.

Meta spokesperson Tracy Clayton told the BBC that "the onus is ultimately on individual people to not actively exploit" the technology, per BBC. Privacy advocates and legal experts have characterized that framing as treating bystander harm as a user-conduct problem rather than a product-design one a distinction that becomes harder to sustain if the next version of the product can identify the people being recorded.

That is why the documented bystander harm matters to the face-recognition story. The current product creates a condition: people in the wearer's environment have no knowledge they're being recorded and no recourse. The proposed upgrade would change what a wearer can do with that condition.

What reports about Meta smart glasses face-recognition code actually say

The New York Times reported in early February 2026 that Meta was considering adding facial-recognition technology to its smart glasses, targeting a year-end launch, based on an internal company document. A Journal of High Technology Law summary of that reporting, published last month, describes the feature internally called "Name Tag" as allowing wearers to identify individuals and receive real-time information about them through Meta's AI. The memo described the feature as a product differentiator. It also described the anticipated launch window as advantageous because civil society groups "would have their resources focused on other concerns," per EFF in February.

The Journal of High Technology Law notes that the feature would not enable unrestricted facial recognition, though the specific limitations described in the memo have not been independently verified.

The legal problem is structural. The current lawsuit concerns footage that buyers actively submitted to Meta AI being reviewed without adequate notice at least that is the allegation. Facial recognition would generate biometric identity data on people who made no product choice, are not Meta users, and have no mechanism to refuse. The EFF has stated this directly: Meta cannot obtain consent from bystanders because the architecture makes it impossible, per EFF. Dozens of state biometric laws, including Illinois's BIPA which requires affirmative written consent before facial geometry can be collected or stored are structurally incompatible with scanning unaware strangers in real time, per the Journal of High Technology Law.

After the reporting became public, EPIC and EFF urged privacy regulators and state attorneys general to investigate before any launch, warning the feature could "significantly expand surveillance risks and threaten civil liberties," per the Journal of High Technology Law. Both organizations also wrote to the FTC, citing Meta's recent security record, though the Journal of High Technology Law's account of the FTC letters contains what appears to be a transcription error in the organization name. Those letters have traction because of what Meta's facial-recognition history actually looks like: a $5 billion FTC settlement in 2019 that included allegations about deceptive face-recognition settings, a $650 million Illinois BIPA class-action settlement in 2021, and a $1.4 billion Texas settlement in 2024 roughly $7 billion in total before Meta shut down its platform-wide face-scanning system and deleted more than one billion face templates, per EFF. That is the record regulators are reading when they receive letters asking them to act pre-launch.

What happens next

The initial case management conference in Bartone and Canu v. Meta is scheduled for tomorrow, June 5. Meta had not filed a formal defense as of this week and has reportedly ended its contract with Sama, the Nairobi-based annotation firm at the center of the Swedish investigation, per MediaLaws. That contract termination removes the specific subcontractor whose workers described the most vivid footage, but it is not an admission of liability.

Three developments would materially shift the legal and product landscape: Meta directly addressing the gap between its marketing claims and its disclosure practices, rather than pointing to supplemental policy text; a decision to delay or shelve Name Tag in response to regulatory pressure; or a regulator establishing biometric consent requirements specific to wearable cameras before any facial-recognition feature ships. None of those has happened.

If smart glasses sales follow the trajectory researchers project up to 100 million owners in the coming years most people affected by the bystander question will never have bought a pair, per BBC. Apple, Snap, and Google are all developing competing products that will require cameras and AI. Whatever standard the Bartone case establishes for adequate disclosure to buyers, and whatever regulatory floor gets set for bystanders in response to Name Tag, will shape what every product in this category is allowed to do. The case management conference tomorrow is the first procedural step. It is not the last thing worth watching.