Meta Says It Fixed Smart Glasses That Recorded Without the Wake Word

A pair of smart glasses that starts recording on its own is the privacy nightmare the whole category was warned about, and this week it happened to Meta. Owners of the company's AI eyewear discovered that their glasses were capturing voice recordings and generating transcripts without anyone saying the wake phrase, according to tech press reports and TechSpot. One user found the behavior across three separate pairs, spanning both the second- and third-generation models, and other owners confirmed similar experiences in the discussion. Meta has confirmed the issue and says it is fixed, but the company has not explained what caused it. The reports describe a textbook case of Smart Glasses That Recorded on their own, with no wake phrase involved.

The discovery unfolded the way most wearable-software scares do: a user noticed something odd in the companion app. Saved audio recordings and text transcripts of conversations were appearing in the voice history even though the "Hey Meta" trigger had not been spoken, and in some cases the "respond without wake word" setting was switched off entirely. The white LED on the inner frame, the tiny indicator meant to signal an active recording, was sometimes lighting up during ordinary conversation. The telltale clue, reported widely by the owners who checked, was that the spurious recordings appeared to follow a firmware update.

Meta's response was brief. "This issue has been resolved. People can manage, review and delete voice logs of their interactions with Meta AI via settings at any time," a Meta spokesperson said in a statement to Android Authority. The statement did not say how many users were affected, whether recordings were limited to audio, or what the root cause was. When Meta support was first contacted by the user who raised the alarm, the advice was to disable microphone and camera permissions, turn off wear detection, delete voice history, and switch off automatic updates, steps the user said would gut the product's core features without explaining the problem.

What owners should check right now

If you own a pair, the practical move is to audit your own voice history. Open the Meta AI app on your phone, go to Glasses, then Device settings, tap the name of your glasses, and look under Glasses privacy for the voice activity log, according to instructions summarized from TechRadar's coverage of the reports. Any recording you did not trigger can be reviewed and deleted there. Make sure both the glasses and the app are running the latest software, since the fix most likely arrived through an update, and keep an eye on that inner-frame LED. If it lights up when you have not given a command, report it to Meta support. Disabling voice activation for the onboard assistant also appears to stop the behavior, for those who would rather be cautious while the details remain thin.

The incident lands in a week when always-listening wearables are already under a microscope. Amazon has just launched its Bee wearables, which record conversations all day as a stated feature. The difference is consent: Bee buyers sign up for continuous capture, while the eyewear owners in this story say they never did. That distinction matters for regulators too. In September, the Dutch data protection authority warned that filming identifiable people without consent is not permitted, and noted that AI functions processing camera images and ambient sound on smart glasses raise privacy questions, a warning issued as the camera-free audio-only variant of the glasses launched.

Why this one hits the category hard

The timing could hardly be worse for trust in camera glasses. Earlier this year, an investigation reported that videos recorded by the glasses were being sent to offshore workers for data labeling used to train AI models, including intimate footage that users likely never knew third parties had seen. Layer that history under this week's unprompted recordings, and the pattern starts to look less like isolated glitches and more like a product whose privacy story keeps getting written by its failures. Meta has tried to shore up the hardware side: the company rolled out a mandatory update that disables the camera if someone tampers with the LED privacy light, closing off the most obvious way to defeat the recording indicator.

The counterpoint is that the system, such as it is, worked the way it was designed to at the most basic level. The LED did light up during the unintended recordings, which is exactly what the indicator is for. The bug was in the triggering, not in a hidden microphone bypassing every safeguard. Meta confirmed and patched the problem within days of the reports surfacing, and users retain the ability to review and delete their voice data. The bigger bet behind the hardware is something GenZ NewZ examined at the September launch: Hot Take: Meta wearable AI Is Zuckerberg's Biggest Bet. For a category that lives or dies on the question of whether bystanders can trust the little white light, that is the thin line between a scandal and a scare.

Still, the unresolved questions are the ones that matter most. Meta has not said what in the software caused the assistant to wake itself, how long the bug was live, or how many pairs were affected. A class-action possibility has already been floated by the user who documented the issue, according to TechSpot. And the broader context is not going away: the Dutch regulator's warning, the labeling investigation, and the arrival of gadgets like Bee that normalize all-day recording mean every new incident lands on a public that is already primed to distrust the form factor. The glasses may have stopped recording on their own. The conversation about whether they should be recording at all has only gotten louder.