Meta's glasses quietly gained face-recognition code on millions of phones

Meta’s smart glasses platform appears to have an unreleased face-recognition system already baked into it, according to code reviewed by WIRED. The system is designed to identify people by using biometric data stored on users’ phones, which means the capability is not just about the glasses themselves but about a linked phone-based pipeline that can connect what the wearer sees with who the wearer is looking at. That is the real headline: the tech was found in code, not in a public product announcement, and it points to face recognition sitting inside Meta’s wearable stack before most users would have any reason to know it exists.

That matters because face recognition is one of the most sensitive forms of consumer tech in the market. Unlike a search query or a location ping, biometric identification can turn a casual interaction into a data event with identity attached. In practical terms, if a smart glasses platform can identify people through biometric data stored on a phone, the product stops being just a camera, display, or assistant and starts inching into surveillance-adjacent territory. Even when no public launch has happened, the mere presence of the code suggests that the technical path has been laid. For anyone building wearables, AI devices, or camera-enabled consumer products, that is the kind of thing privacy teams, product leaders, and legal counsel usually want to see before it reaches millions of hands.

WIRED’s reporting is tight on what has been confirmed and careful on what has not. The article says code reviewed by WIRED uncovered the unreleased system, and that it is designed to identify people via biometric data stored on users’ phones. It does not say the feature has been rolled out publicly, nor does it say Meta has announced it. That distinction matters. In tech, code can be a roadmap, a test, or a nearly finished feature sitting in plain sight. But once a capability like face recognition exists inside a platform, it becomes much easier to imagine product teams turning it on, iterating on it, or tying it to another feature later. That is why companies often keep these systems tightly controlled: once the code exists, the governance burden arrives early.

For Meta, the story lands in a broader fight over how far smart glasses and other AI wearables should be allowed to go. Glasses are appealing because they are worn, always on, and close to real life. That makes them useful and powerful. It also makes them politically and socially sticky. A device that can observe the world and identify people brings together several things regulators and consumers tend to scrutinize separately: cameras, biometric data, device permissions, and the permanence of identity data. If the biometric data is stored on a user’s phone, that suggests the phone is not just a companion device but a key part of the identity system. For device makers, that creates a new dependency. For users, it can create a new trust question: who can access the data, and under what conditions?

There is also a second-order implication for boardrooms far beyond Meta. The last few years have made one thing obvious: the fastest product features are often the ones that create the slowest policy headaches. If a company is building a wearable or AI-powered consumer product, code like this is not just an engineering detail. It becomes a question of consent design, disclosure, platform risk, app store review, and future regulatory scrutiny. A feature that can identify people from biometric data could trigger a very different conversation than one that merely recognizes objects or reads text. And because the system is embedded in the smart glasses platform, executives should read this as a signal that the platform architecture itself can determine the legal and reputational blast radius later.

The broader market context is simple: smart glasses are moving from novelty to platform, and platforms tend to absorb more sensitive capabilities over time. That is usually how consumer tech expands. First the device does the obvious thing. Then it adds assistant features. Then it adds personalization. Then it starts learning more about the environment, the user, and the people around the user. Each step feels incremental until one day the product has crossed into a category that invites heavier scrutiny. Face recognition is one of those line-crossing capabilities because it is easy to explain and hard to defend without careful controls. It can make a device feel magical. It can also make the device feel invasive in a hurry.

For executives, the practical lesson is not that every unreleased feature is a scandal. It is that hidden capabilities in code have a way of becoming governance problems before they become revenue. Competitors in wearables, computer vision, and consumer AI should assume that anything touching biometric identification will be judged not only by what it does, but by how quietly it got there. Meta’s situation shows why product, privacy, and public policy teams need to be in the room early, not after code review turns up a surprise. In a category where trust is part of the product, the difference between a feature and a liability can be one line of code too many.