--- title: "The Light Was On: Meta’s Ray-Ban Glasses Recorded When No One Called Them" description: "A bug let the glasses record audio without their wake word. Meta says it is fixed but has not said what caused it" author: "Jan Sokol" published: 2026-10-16T19:11:50Z modified: 2026-10-17T06:38:36Z url: https://rews.cc/a/the-light-was-on-meta-s-ray-ban-glasses-recorded-when-no-one-e947fb language: en tags: ["privacy", "meta", "ray-ban", "surveillance", "ai", "tech"] publisher: "Rews (https://rews.cc)" --- # The Light Was On: Meta’s Ray-Ban Glasses Recorded When No One Called Them *A bug let the glasses record audio without their wake word. Meta says it is fixed but has not said what caused it* By Jan Sokol · October 16, 2026 · https://rews.cc/a/the-light-was-on-meta-s-ray-ban-glasses-recorded-when-no-one-e947fb ## In brief - A bug caused Ray-Ban Meta glasses to record audio without the wearer meaning to, even with the recording light on - Users said the glasses activated Meta AI without the “Hey Meta” wake word, in some cases with the wake word turned off - Meta told Android Authority the issue was patched but did not explain the cause or the fix - A Swedish report quoted Sama workers in Kenya who said they had seen footage apparently recorded without people’s knowledge - The report led to a proposed US class action and a Texas investigation, and Norway is considering a temporary ban For years Meta’s answer to people who feared its glasses was a small light. When a pair of Ray-Ban Metas records sound or images, a light on the frame turns on. In the company’s account, that light is the warning and the courtesy, a kind of agreement between the person wearing the camera and the people it points at. This week the light turned on, and it guaranteed nothing. As [Scharon Harding reported for *Ars Technica*](https://arstechnica.com/gadgets/2026/10/ray-ban-meta-smart-glasses-found-recording-audio-without-wearers-consent/), a bug caused the glasses to record without their wearers meaning them to, and the indicator was lit while it happened. The fear that has followed the device since its launch, the fear of being recorded without consent, stopped being hypothetical. ## Without the magic words The glasses are supposed to wait for a summons. The owner says “Hey Meta” and the assistant wakes up. According to [Android Headlines](https://www.androidheadlines.com/2026/10/meta-fixes-ray-ban-smart-glasses-bug-after-reports-of-unexpected-recording.html), which cited 9to5Google’s original report, several users found that their glasses were starting Meta AI and recording audio on their own. Nobody had said the wake word, and in some cases the wake word had been switched off altogether. One user said the same thing happened on three separate pairs. Some called it a “privacy nightmare.” The recordings were audio only. The cameras were not involved. So the machine heard a command that no one gave, and it obeyed. ![Dark blue glasses connect to a stylized ear with cream sound waves on a teal background.](https://rews.cc/img/88161e7142f0414e3e6e61528ce6e86a9c659461.webp) ## The statement Meta told Android Authority that the problem has been patched. It did not say what caused it or how it was fixed. What it offered instead was a reminder of the controls available to the owner: > People can manage, review, and delete voice logs of their interactions with Meta AI via settings at any time. The sentence is worth reading slowly. It addresses the owner, the one person the bug did nothing to protect, and it puts the job of cleaning up on that owner. It says nothing to the other people in the room, whose voices may have been captured and who have no settings to adjust. Large institutions often answer a failure by describing the tools they have handed to someone else. That is my reading, not anything Meta has said, but the statement leaves a lot of room for it. ## What others have seen This latest bug arrived on top of a story that was already uncomfortable. Earlier this year Swedish journalists interviewed more than 30 workers at Sama, a data annotation company headquartered in Kenya. These are the people paid to look at what the machines collect. The workers said they had viewed footage from Ray-Ban Metas that seemed to have been shot without the wearer, or anyone else in the frame, knowing. The report described a “stream of privacy-sensitive data that is fed straight into the tech giant’s systems,” as rendered by machine translation. Among the private moments the annotators said they saw was a glasses owner’s partner walking out of a bathroom naked. ![A seated person examines a filmstrip showing repeated views of an open bathroom door, a red towel, and a bathtub against a teal background.](https://rews.cc/img/5ee9fc02feb2beb31afe7c79538e9206f910b124.webp) “Meta has that type of content in its databases. People can record themselves in the wrong way and not even know what they are recording,” one anonymous Sama employee told the reporters. The bug lends some weight to that account. The tidy picture of these devices has a careful owner in control and a bad actor as the only threat. The workers in Kenya describe something messier: owners who mean no harm, recording things they never intended, with the footage going on to strangers on another continent whose job is to label it. ## The response so far In the United States, the Swedish report led to a lawsuit seeking class action certification and to an investigation by Texas Attorney General Ken Paxton. Meta has also patched the device itself once already. It updated the glasses so they will not record if the indicator light is covered. *Ars Technica* dates that change to August, while Android Headlines describes a July update that disables the camera when the privacy light is tampered with. That safeguard was aimed at the deliberate abuser, the person who puts tape over the light. It did nothing about a microphone that switched itself on while the light was shining. ![A cream padlock with a black microphone symbol and yellow indicator sits on a blue background, with three curved sound waves to its right.](https://rews.cc/img/49cdeceadd8192a1ab578b43b1a46f4441eb1813.webp) Outside the company, people have started taking their own precautions. Apps have been built to detect smart glasses nearby. In the Netherlands, [retailers suspended sales](https://rews.cc/a/dutch-retailers-suspend-sales-of-meta-s-smart-glasses-as-fil-540ee9) of Meta’s glasses as the backlash over filming grew, and [governments have begun to pay attention](https://rews.cc/a/governments-are-starting-to-notice-the-camera-on-your-face-453c99) to cameras worn on the face. Lawmakers have proposed legislation, and Norway is considering a temporary ban. Privacy can be breached in more than one way. There are technical faults like this week’s. There is data shared without the user’s agreement. There is design that makes accidents easy. The worry has been building for some time, alongside reporting that [facial recognition has spread](https://rews.cc/a/kashmir-hill-says-facial-recognition-spread-to-glasses-apps-84f766) to glasses and apps. Meta says this bug is closed. It has not said how many people were affected or how long the problem lasted. It has not said what happens to the audio already captured, beyond the reminder that owners can delete their logs. The people who were recorded by accident in kitchens and on streets have no settings to open. Their conversations may still be stored on Meta’s servers until an owner, who may not know they exist, decides to delete them.