Meta AI smart glasses facial-recognition code

Summary

On June 4, 2026, WIRED reported finding code for an unreleased facial-recognition feature, named NameTag internally, inside the Meta AI app that pairs with the company's smart glasses. A security researcher who reviewed the code said it was dormant. Meta removed the code in an app update within days and said no final decision had been made about launching the feature.

What happened

• WIRED reported that the Meta AI app, which connects to Meta's Ray-Ban smart glasses, contained a facial-recognition system labeled NameTag.
• WIRED reported that the code referenced three machine-learning models: one to detect faces, one to crop them, and one to convert them into biometric signatures.
• A security researcher who examined the code said the system was dormant and was not sending biometric data to Meta's servers.
• WIRED reported that the dormant code shipped on more than 50 million devices with the app installed.
• EFF's Threat Lab said it verified the code through static analysis.
• Meta later removed the facial-recognition code, the models, and the alerting logic in an app update.

Timeline

• 2026-06-04 -- WIRED publishes its report on the NameTag code in the Meta AI app.
• 2026-06-05 -- Meta releases an app update removing the facial-recognition code and models, per EFF.
• 2026-06-08 -- Press coverage confirms the removal; Meta states no final decision had been made about launching the feature.

What the vendor has confirmed

Meta acknowledged removing the code and said no final decision had been made about whether the NameTag feature would launch. The company did not publish a postmortem or describe what biometric data, if any, it processed during internal testing.

What remains unclear

• Meta has not disclosed what biometric data, if any, it processed while testing the feature.
• Meta has not said whether the NameTag system will be enabled in a future release.
• Whether the dormant code could have been activated without a further app update has not been established.

Broader context

Biometric identification built into a consumer wearable raises a distinct privacy question, because the faces captured usually belong to bystanders who never agreed to be scanned. Code shipped in a dormant state also shows that the capability inside an installed application can exceed the features a user sees, since functions can be staged ahead of any public launch.

Sources

• EFF Threat Lab analysis (research)
• Engadget coverage (press)
• Rappler coverage (press)