Meta Tests ‘Super Sensing’ AI Glasses That Record Every Moment

Meta is testing a prototype pair of “super sensing” AI glasses that would keep their cameras and microphones running for hours at a time, building a continuous visual and audio memory of a wearer’s day. The Financial Times reported the prototype on Wednesday, drawing on details first published by The Information and reviewed by multiple outlets since. The test sits at the seam where Meta’s consumer AI meets one of the hardest questions in wearable computing, which is what the small white light on the frame of a pair of smart glasses is actually for.

What ‘Super Sensing’ Actually Does

The feature set Meta is building under the super sensing name is a longer-running version of Live AI, the always-listening assistant that already ships on current Ray-Ban Meta glasses. The current device can keep Live AI active in the background for roughly 30 minutes before the battery gives out. The new mode is engineered to run for hours, with the cameras and microphones staying on the whole time, so the assistant can keep a continuous memory of what the wearer sees and hears.

The super sensing feature set covers a small set of capabilities the reporting names explicitly:

• Continuous camera capture, taking photos every few seconds without a user prompt.
• Continuous audio capture, so the assistant can listen to the room as well as look at it.
• A facial recognition mode the company is “exploring,” which could surface the name of someone the wearer is talking to from stored information.
• Hours-long background sessions, the defining change versus today’s 30-minute ceiling.

The user experience is meant to feel like a passive memory. The Information’s examples, as relayed by UploadVR, are concrete: a reminder to grab keys the wearer left behind, or a nudge to stop in a store for ingredients the assistant heard the wearer mention earlier. The glasses would not be used only when the wearer remembers to ask. They would already be listening.

The LED Question Is the Real Fight

Inside Meta, the question sitting on top of every other super sensing question is whether the small white LED that lights up when the glasses are recording should stay on in the new mode. The Verge reports that Mark Zuckerberg has personally questioned whether the indicator could stay off during always-on Live AI sessions, and that Meta is “weighing the idea.” The current smart glasses do light the LED when the camera is active, whether for capture or for AI use, and the design is the only visual signal a bystander gets.

Meta’s own 2025 policy document, written when the company first turned off the LED for routine AI queries, explains the trade-off it accepted at the time. “If the LED light were to flash continuously over extended periods, for instance every time an AI interaction occurs, people might become desensitized, thereby reducing users’ awareness of when photos or videos are being captured.” The company added that it would “implement other measures to protect people’s privacy, such as removing key identifiable information.” The same logic, applied to a mode that runs for hours rather than for the occasional question, lands differently.

Zuckerberg hinted at the technology during the company’s first-quarter earnings call, stating his vision for glasses to evolve from merely “answering questions” to becoming “a personal assistant that stays with you all day, helping you remember things and achieve your goals.”

The Mashable report notes that Meta is “considering an indicator that would let people know the super-sensing feature is active,” a softer formulation than the bright always-on LED on the frame today. What kind of indicator, where it would sit, and who would see it are all still open. The decision the company lands on is the consequential one, because the LED is the entire social contract between the wearer and the room.

How Meta Got Permission to Build It

Super sensing is not a product that arrived in a vacuum. The reporting chains the prototype to two earlier decisions the company made quietly over the past year. The first is an internal shift in how Meta weighs privacy risk. The Information previously reported that Meta’s privacy, integrity, and legal teams now have less authority to block ideas, while product teams have more autonomy on these issues. The shift is the institutional backdrop for a feature that other teams inside Meta would once have slowed down.

The second is a public change in the default settings on the existing glasses. In April 2025, Meta updated the privacy policy for the Ray-Ban Meta line in two ways the Mixed News report walks through. Meta AI is now enabled by default, although users can still turn it off. And it is no longer possible to opt out of allowing Meta to store voice recordings to improve its AI products, although the recordings can be manually deleted. Each move is small on its own. Together they reset the baseline the user agrees to, from a privacy-on experience to a privacy-off experience.

Facial recognition makes the chain longer. Mashable reports that Meta originally scrapped facial recognition for the first generation of the Ray-Ban Meta AI glasses over ethical concerns, the same decision Facebook made in 2021 when it shut down the broader facial recognition system on the social network. The current reboot, the outlet says, is credited to a more business-friendly Federal Trade Commission under the current administration, a regulatory climate in which Meta feels it can take the idea back off the shelf.

The Two Devices Behind the Code Names

Super sensing is being built for two specific next-generation devices, not as a software update to current hardware. UploadVR, reporting on The Information’s piece, names them as Aperol, a sunglasses design, and Bellini, a prescription pair. The same report pegs both to a launch window of late 2026 or early 2027, and notes that production capacity for the Ray-Ban Meta line is being lifted to 10 million units by the end of 2026 from a base of 2 million units sold as of February 2026.

Both Aperol and Bellini sit inside a wider Meta glasses roadmap that now runs through 2027:

Current Ray-Ban Meta and Oakley Meta glasses start at $299 on Meta’s AI glasses product page. The super sensing feature is described as opt-in for the wearer, and the company is also said to be working on a pair of AI earphones with embedded cameras and the same sensing stack, a second body of hardware that would extend the same always-on question to the ears.

A Legal Landscape That Has Noticed

The legal system is already circling. Recording audio of third parties without their consent is illegal in multiple U.S. states, a fact the Moomoo summary of the FT reporting pulls out, and the EU’s General Data Protection Regulation was written before always-on cameras on faces became a consumer product. The structural problem is that the bystander has no way to consent to being filmed by a device the wearer is wearing down the street, in a bar, or at a work meeting.

More than 70 advocacy groups have written to Meta asking the company to abandon the facial recognition plans, and the academic critique is pointed. Woodrow Hartzog, a law professor at Boston University School of Law, told the Financial Times that the existing framework does not match the device.

“No existing law comprehensively addresses the myriad ways these tools can pose risks during their design and manufacturing. Legislators must take this issue seriously and update the legal framework to reflect the realities posed by devices that are always on and constantly observing.”

The Mashable reporting adds a regulator-specific beat: the same loosening Meta reads as license to ship is the loosening the advocacy groups and the academics say is the reason new rules are needed. The two readings of the same shift put the prototype squarely in the middle of a policy fight Meta did not start.

The Bystander’s Side of the Trade

The opt-in question is the one that most decides what super sensing means in practice. The Verge and the Mixed News report both state that the mode is opt-in for the wearer, with Mixed News specifying that it “will likely be turned off by default and will need to be turned on.” Neither report describes a parallel opt-out path for the people the glasses are pointed at. The bystander is the part of the social contract that is not in the consent flow, and the design reflects that. The wearer’s on switch is bystanders’ off switch they do not have.

Eric Abbruzzese, a research director at market intelligence firm ABI, told CNET that the no-display form factor is “an ideal testing ground both for the AI platforms itself and how it may interact or be used on smart glasses devices.” The glasses are a real product today, sold in volume, and the AI running on them is the consumer install base for whatever super sensing turns into. A feature that asks bystanders for nothing while giving the wearer hours of continuous capture is a quiet rebalancing of who a public space belongs to, and the FT and The Information reporting suggests the rebalancing is what Meta is testing in market rather than in committee.

What Lands When, and What Could Still Break

The launch window the leaks point to is late 2026 or early 2027, the same window the uploadvr technical breakdown lays out. Two things have to land for the prototype to become a product. Meta has to confirm the feature, and the company has to decide what the LED does when super sensing is running. As of the FT report, Meta has not commented directly on the super sensing prototype, and the CNET request for comment went unanswered. Hannah Murphy, one of the FT reporters on the story, noted on the FT News Briefing podcast that the team has “not heard from Meta yet directly on the super-sensing glasses.” That silence is itself a signal: the product is in the testing phase, not the announcement phase.

The key figures the leaks have put on the table, all of them from the Information-led report on the next-gen smart glasses:

• 2 million Ray-Ban Meta units sold as of February 2026
• 10 million target production capacity by end of 2026
• Live AI today: roughly 30 minutes of continuous use
• Super sensing target: hours of continuous use
• Ray-Ban Meta starting price: $299

Two near-term risks sit on top of the launch plan. The first is facial recognition: The Information calls it an “exploration,” not a feature, and Meta previously killed the idea under the same ethics review the company has since pared back. The second is the LED policy itself, where a single change in the indicator behavior would reset the public conversation. Hypernova, the 2025 HUD glasses Meta plans to ship first, already carries a “slight risk” of slipping past its October target, the same Bloomberg report notes, and the roadmap is fluid enough that an Aperol or Bellini delay is the kind of slip the company has made on past products. A super sensing launch is most likely to land inside the late 2026 or early 2027 window the leaks describe, and to land looking different in product than it does in the FT report.

Frequently Asked Questions

What is Meta’s “super sensing” mode?

Super sensing is the internal name for an always-on Live AI mode Meta is building for its next-generation smart glasses, in which the cameras and microphones stay active for hours at a time. The assistant can take photos every few seconds, record snippets of audio, and surface reminders based on what it has seen or heard.

When will Meta’s super sensing glasses launch?

The Information points to a launch window of late 2026 or early 2027 for the two next-generation devices that will support the mode, codenamed Aperol and Bellini. Meta has not publicly confirmed the timing.

Will the recording LED stay on?

Current Ray-Ban Meta glasses light a white LED whenever the camera is active, including during AI use. The Information reports that Mark Zuckerberg has questioned whether the LED could stay off in super sensing mode, and that Meta is “weighing the idea.”

Is facial recognition coming to Meta glasses?

The Information reports that Meta is “exploring” a facial recognition capability that could surface the name of someone the wearer is talking to. Meta scrapped the same feature for the first generation of the Ray-Ban Meta line over ethical concerns, and more than 70 advocacy groups have asked the company to abandon the new plan.

Can bystanders opt out of being recorded?

The reporting describes super sensing as opt-in for the wearer, and the Mixed News summary says the mode will likely be off by default. None of the leaks describe a parallel opt-out path for people being filmed or recorded, and recording third parties without consent is illegal in multiple U.S. states.