|
Getting your Trinity Audio player ready...
|
A pair of smart glasses is supposed to stay quiet until you speak to it. This week, a group of Ray-Ban Meta owners said theirs did not. They found saved audio clips and written transcripts of everyday conversations in the Meta AI app, even though they had never said “Hey Meta.” Meta says the problem is resolved, but it has not said what caused it.
The complaint began with a post on Reddit from an owner who noticed unfamiliar entries in their voice history. The entries came from ordinary chats with other people, not from requests made to the assistant. The setting that lets Meta AI respond without the wake phrase was switched off. The owner also said the small white light on the glasses sometimes came on while they were talking normally, a sign that the microphone had been triggered.
What made the report harder to dismiss was its spread. The same behavior was described on three separate pairs, covering both second and third-generation models. The owner said it started after a firmware update. One misbehaving pair might point to a faulty unit, but identical symptoms across several devices and two hardware generations make a software cause the more natural suspect. That remains an inference from the pattern, since Meta has not confirmed it.
The owner’s experience with customer support added to the frustration. The advice reportedly given was to turn off microphone and camera permissions, disable wear detection, clear voice history and stop automatic updates. For a product sold on hands-free voice and camera features, that amounts to switching off most of what it does. The owner also said they were speaking to a lawyer about a possible group legal claim if others had seen the same thing. No case has been reported, so that remains an intention rather than a development.
When Meta was asked about the reports, it said the issue “resolved” and pointed users to the settings where they can manage, review and delete their voice logs. It did not say whether the firmware update was involved, how many devices were affected, or what became of audio captured during the unintended activations. In other words, the company has offered reassurance without an explanation.
There are limits to what can be said with confidence. The accounts so far come from owners posting online, and a handful of others in the replies said they had noticed similar unexplained activations. There is no independent testing showing the fault was widespread, and nothing outside Meta’s own word confirms the fix works. Both the problem and the cure are, for now, unverified by anyone other than the people involved.
Even so, the episode matters because of how these glasses are marketed. Meta’s public privacy messaging says the glasses become active after the wake phrase or a touch on the frame. Ray-Ban’s help material describes the front light as a signal that voice commands are on or that the wake word has been heard, and says the system is designed to ignore speech not meant for it. If transcripts can appear without a deliberate command, the difference between that promise and the owners’ experience is the real story, whatever the bug turns out to be.
Storage raises a second concern. Meta’s privacy notice indicates that voice interactions with its assistant can be kept for up to a year unless the user removes them. Recordings the owner never intended to make could therefore sit in an account for months without being noticed. Checking the voice history is a small task with a real payoff.
The glasses were already under scrutiny before this week. Earlier reporting described contract workers reviewing footage from the devices that included deeply private scenes, and Meta has acknowledged that some user content is shared with contractors to improve its products, with protections applied. Separately, wearers found ways to hide the recording light, which led Meta to issue updates. The most recent change makes the camera stop working if the light is covered during a recording, a fix announced by Meta’s head of wearables on Threads. That fix concerns video and the people being filmed, so it is a different matter from the audio fault now being reported. Still, each episode chips at the same foundation, which is the public’s willingness to trust a small light and a company’s assurances.
Governments and institutions are reacting as well. Norway is reported to be considering a temporary ban on smart glasses in parks, beaches and schools, and courtrooms in the United States and elsewhere already restrict them. Those moves target filming more than wake-word faults, but they show how uneasy people are about wearable devices that can sense their surroundings without an obvious sign.
There is also a conceptual problem that the industry has yet to settle. AI wearables can listen, transcribe and analyze without necessarily creating the kind of audio or video file people picture when they hear the word “recording.” When a transcript appears in a user’s history after a casual conversation, that technical distinction means little to the wearer or to the person they were talking to. What matters is whether anything was captured, who can see it and how long it stays.
For people who own these glasses, the response can be calm and practical rather than alarmed. Open the Meta AI app and go through your voice activity history. Delete anything you did not mean to save. Confirm that the “Respond without Hey Meta” option is set the way you want, and install any waiting firmware or app updates, since that is how Meta says the fix is delivered. Product details and settings are on Meta’s smart glasses page, and Ray-Ban publishes its own guidance for the Meta AI range. If you wear the glasses around colleagues, clients or friends, it is considerate to say when voice features are on, especially in private or professional settings.
Several questions remain open. Meta has not named the cause, has not confirmed a link to the firmware update, and has not said how many people were affected or what happens to audio captured by mistake. Until it does, a single word saying the problem is resolved is a claim, not an explanation. Readers who follow privacy and consumer technology can find more coverage on Business Tech.
