After WIRED documented a face recognition system called NameTag embedded in the Meta AI companion app, Meta removed the code a day later. The prior denials and the unanswered questions about data retention and timing survived.
Meta's public position on facial recognition in its smart glasses was that the feature did not exist. Andy Stone, the company's vice president of communications, told WIRED in January that the system "does not exist." Andrew Bosworth, Meta's chief technology officer, called WIRED's reporting on the feature "incredibly misleading" and "absolutely dishonest."
One day after WIRED published its investigation on Thursday, the latest version of the Meta AI companion app stopped containing the code libraries that powered that system. The app, which is installed on more than 50 million phones and serves as the companion software for Meta's line of smart glasses, no longer ships with the unactivated face-recognition components WIRED had documented by comparing successive app versions.
The contradiction shows up in the code itself. The version of the Meta AI app available on the day WIRED published its story contained several code libraries explicitly named for face recognition. The release Meta pushed the following Friday contained none of them. Meta did not announce the change. The removal was visible only to anyone who, like WIRED, compared the two builds.
The feature, internally called NameTag, was designed to convert faces captured by Meta's smart glasses into unique biometric signatures and compare them against a database. WIRED's investigation found that NameTag was never publicly enabled. But the code that would have powered it sat inside the Meta AI app, distributed to tens of millions of users, for months. The face-recognition machinery was already present in a January build of the app, predating Meta's public denial.
Stone's posture shifted once the removal became visible. He told WIRED on Monday that NameTag is "purely exploratory" and that "no final decision has been made on what to do here, if anything." That language is closer to Meta's earlier public framing of the feature as a research project, and a long way from "the feature does not exist."
The shift is the story. WIRED's original reporting established that the system was present in the app and that the prior denials were wrong. Meta's response was to call the reporting dishonest. Meta's next observable action was to remove the code. Those two things cannot both be true.
What survives the code removal is the accountability surface the contradiction exposed. WIRED's investigation documented that faces the system failed to recognize were cropped from the user's view, indexed, and stored locally on the device for future processing. WIRED also reported, citing an internal Meta memo first surfaced by The New York Times, that Meta had discussed launching the feature during a "dynamic political environment," when privacy and civil-liberties advocates would be distracted. Meta declined to answer ten specific questions WIRED submitted about data retention, the face-profile database, and the January build of the app that already contained the face-recognition machinery.
Pulling the code does not answer any of those questions. It does not say what happened to the faceprints and indexed images the system was already configured to store. It does not say whether copies of the cropped faces were transmitted off device, retained in telemetry, or shared with the face-profile database the code was built to query. It does not say whether the January build of the app, which predated Meta's public denial, ever ran the face-recognition components against a real user's face.
It also does not address the strategic timing question the internal memo raised. A company that publicly denies a feature exists, calls the press that documented it dishonest, and then quietly removes the code a day later has not yet explained why it built, embedded at scale, and defended the feature in the first place. Stone's "no final decision" framing leaves room for Meta to re-enable the system, in some form, once attention moves on.
Meta's smart-glasses business depends on social permission to put cameras on faces in public. The corporate response to the first serious investigation of NameTag was to deny the system existed, to attack the credibility of the outlet that found it, and then to take the system out of the app. Each step was an attempt to keep the contradiction from becoming a public fact. The code diff is the fact.
What to watch: whether Meta files a public statement, blog post, or developer note explaining the removal; whether any of the unanswered retention questions surface in regulatory filings, congressional testimony, or Meta's own privacy disclosures; and whether the face-recognition components reappear in a future build under a different name.