Are Meta AI glasses private? Voice recordings, camera AI, and the LED

The Morse teamPublished October 26, 202610 min read

As of October 2026, Meta AI glasses are private for the photos and videos you take for your gallery, which stay on the glasses until you import them, but not for what you say or show to Meta AI. Ray-Ban Meta and Oakley Meta glasses store your voice requests on Meta's servers by default for up to a year, with no switch to stop it, and images Meta AI takes to answer you can be stored, used for training, and reviewed by people unless you turn off a setting Meta added in September 2026.

We read Meta's AI glasses voice notice, its supplemental privacy policy and terms for the glasses, Meta's AI terms, seven Meta help pages for the glasses, Meta's responsible innovation page, Meta's July 2026 glasses FAQ, and Meta's September 2026 engineering post on October 26, 2026, along with the Texas attorney general's May 2026 release and the federal court docket. Meta deserves credit in two places. Its voice notice gives exact retention periods, and from the second generation on, its glasses disable the camera when someone covers the recording light.

Meta AI glasses privacy at a glance

These are the US rules for Ray-Ban Meta and Oakley Meta glasses, from Meta's own pages, read October 26, 2026.

Do Meta AI glasses record your voice?

Yes, every voice request to Meta AI on Ray-Ban Meta and Oakley Meta glasses is recorded and kept. Meta's voice notice, effective July 22, 2025, says "Text transcripts and audio recordings of your voice interactions are stored by default to help improve Meta's products" (Meta). Meta's help page says "The option to disable voice recordings storage is no longer available, but you can delete recordings" (Meta). TechCrunch reported on April 30, 2025 that Meta emailed owners about the change the day before (TechCrunch).

Meta's first-generation Ray-Ban Stories still have the switch. The same voice notice says Ray-Ban Stories owners "can turn voice storage off" in settings, and their recordings are then deleted after processing (Meta).

A recording can pick up more than you. Meta's supplemental privacy policy, effective June 29, 2026, says voice services process "any background sound that can be heard once voice services are enabled" (Meta). The glasses also keep listening after you finish. Meta says Respond without "Hey Meta" "is on by default," so you can ask a follow-up without the wake word (Meta).

How long does Meta keep Ray-Ban Meta voice recordings?

Meta keeps Ray-Ban Meta and Oakley Meta voice recordings for up to one year. Its voice notice says "Voice transcripts and stored audio recordings are otherwise stored for up to one year to help improve Meta's products" (Meta). Meta labels accidental triggers "false wakes" and deletes them "within 90 days of detection" (Meta).

People listen to some of them. Meta's voice notice says stored recordings are processed by machine learning and trained reviewers, and "we protect your privacy by changing the pitch of voice recordings when they are reviewed" (Meta). Meta's help page gives an example of what reviewers check: whether the glasses took a photo when you asked (Meta).

To see and delete them, open the Meta AI app, tap the Glasses icon at the top right, then Device settings, your glasses, Glasses privacy, and Voice activity log (Meta). Meta says deleting a recording also deletes its transcript and the data tied to that interaction (Meta). Deleting doesn't undo training already done. Meta says "If you delete a piece of information a model learned from, the model won't change in that moment" (Meta).

Does Meta use photos from Ray-Ban glasses for AI?

It depends on which photos. Meta's pages split what the camera sees into two kinds, and Meta handles them differently.

Photos and videos you take for your gallery. Meta says these "are stored privately on your glasses" and go to your phone's gallery when you import them (Meta). Meta's supplemental privacy policy says it processes this media when you turn on cloud processing, use Meta AI, or upload it to Facebook or Instagram (Meta). With Cloud media on, photos and videos are sent to Meta's cloud and "temporarily stored in the cloud for 30 days before being automatically deleted" (Meta).

Images and video Meta AI takes to answer you. Meta calls these "visual data," meaning "images and videos from AI experiences on your glasses," such as asking what you're looking at or using Live AI (Meta). Meta's privacy policy says "your information, like Media and audio recordings of your voice, may be used to improve Meta products" (Meta). The same policy says Meta may share "a photo you take to ask Meta AI a question" with third parties as its AI terms describe (Meta). Meta's AI terms, effective May 13, 2026, say review of AI interactions "may be automated or manual (human)" and happens "through third-party vendors in some instances" (Meta).

Meta AI's camera feature is on from the start. Meta's help page says "Meta AI with camera use is already set up," and turning off "Hey Meta" turns off voice access to the camera (Meta). During a Live AI session, Meta says "Your glasses camera and microphone are continuously active" until you pause it (Meta).

How do you stop Meta training on your glasses' camera?

Turn off Store visual data from AI experiences. In the Meta AI app, open your glasses settings, tap Meta AI under Experiences, and switch it off (Meta). Engadget reported the setting on September 23, 2026 (Engadget).

With it off, Meta says "Your visual data will not be used for training or other product improvement, and it will not be stored," and "will not be available for human review" (Meta). With it off, "Some AI features, like visual reminders, may be limited or unavailable."

Switch it off before you use the camera features. Meta's help page says "This change doesn't apply to visual data that was already processed" (Meta). The setting covers images only, and your voice requests stay stored under the voice notice. Articles from 2024 saying there's no way to keep AI photos out of training still ranked for "does meta use photos from ray ban glasses for ai" on October 26, 2026.

Can people tell when Meta glasses are recording?

Sometimes. Meta's FAQ, published July 7, 2026, says "Whenever content is being captured for your gallery, this white light blinks," for a photo briefly and for a video "as long as you are recording," and "The capture LED has no off switch" (Meta). Starting with the second generation, Meta says "the camera is automatically disabled if we detect that the capture LED has been blocked," and it is updating the glasses to do the same when the light is "physically tampered with or destroyed."

Meta's FAQ ties the light to content for your gallery, and its responsible innovation page says the same of photos and videos "for your gallery or streaming" (Meta). None of the Meta glasses pages we read on October 26, 2026 says whether the capture LED lights when Meta AI takes a picture to answer you or during Live AI, and the top results we read for this query describe the light without saying either. Those AI captures are the visual data Meta can store, train on, and send to reviewers. Texas's attorney general alleged on May 20, 2026 that the indicator "is not active during the smart glasses' 'always enabled' mode" (Texas Attorney General). That is an allegation, so far unproven.

Meta puts the bystanders on you. Its supplemental terms say you're responsible for "obtaining any consents from other individuals" where recording or privacy laws require it, and bar you from obscuring features "intended to signal to others that the product is recording" (Meta).

Who has seen footage from Meta glasses?

Contractors may have. Texas's attorney general said on May 20, 2026 that data annotators at Meta's subcontractor Sama in Kenya "have claimed that they have access to video material of users' private moments, such as bathroom visits," and issued Meta a Civil Investigative Demand (Texas Attorney General). Buyers sued too. In re Meta AI Glasses Litigation, case 3:26-cv-01897, was filed on March 4, 2026 in the Northern District of California and is before Judge Edward M. Chen (CourtListener). As of October 26, 2026 the docket shows no final ruling, and the claims are unproven.

Will Private Processing make Meta glasses private?

Meta has announced it but not said when. Meta's engineering team wrote on September 23, 2026 that it is "Bringing Private Processing to Meta AI Glasses," running cloud AI in confidential virtual machines that Meta says it can't read, with the glasses checking each server's attestation before they connect (Meta). The post gives no launch date and doesn't say which features will use it. It's the same enclave approach that Meta's Incognito Chat uses, which we cover in is Meta AI private, and we explain the hardware in what is a trusted execution environment.

How Morse keeps your sensitive questions out of a voice log

Each "Hey Meta" request and each image Meta AI takes goes to Meta's servers under your Meta account. The voice requests are stored by default for up to a year and open to reviewers, and the images can be stored and reviewed unless you switch that off. Morse can't change what your glasses send. For the questions you'd rather not say out loud to Meta AI, about a symptom, a debt, or a person, Morse is a web app where you type them instead. Your chat history is stored only in your browser, encrypted, not on Morse's servers (Morse privacy policy).

• Private, for questions you'd rather no company keep for a year. It's the mode a new chat starts in. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, run here, and Morse has zero data retention agreements with its inference, search, and other providers. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for what Private Processing promises, available today. Your browser checks the enclave's hardware attestation before anything leaves your device, then encrypts the message so only that enclave can read it, and our servers relay text they can't read. The enclaves are run by third-party providers, not by Morse, and the protection rests on the enclave hardware and attestation working as designed. E2EE chats come with Pro, at $19 a month.
• Anonymous, for GPT, Claude, or Gemini without your name on the request. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy, so leave out names you'd rather keep to yourself.

There's no training switch to find, because Morse doesn't train AI on your content, sell your data, or show ads. Morse's servers don't keep messages or log message content. Your history locks when you sign out, and opening it takes two keys, one that never leaves your device and one our server gives you when you sign in. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account deletes your chats, and Morse can't recover them, so keep a copy of anything you need.

So are Meta AI glasses private enough for you?

• Voice: every Meta AI request is stored for up to a year. Delete them in the Voice activity log, and turn off Respond without "Hey Meta" to shorten how long the microphone stays open. On Morse, you type, and the history stays in your browser.
• Camera AI: turn off Store visual data from AI experiences before you ask Meta AI about what you see. Images already processed stay processed.
• Gallery photos: leave Cloud media off if you don't need it, so photos stay on the glasses and your phone.
• People around you: the capture LED shows gallery recording, and the consent rules are yours to follow.
• Sensitive questions: don't ask them out loud. Ask them in Morse, in Private mode, or in E2EE for anything no one should read.

For the rest of Meta's apps, see how to turn off Meta AI and our comparison of which AI companies train on your data.

Try a chat from the Morse home page without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.