Wearable AI was the organising idea of Meta’s Connect 2026 keynote in Menlo Park on Wednesday 23 September. Mark Zuckerberg used the stage to launch new glasses, a lighter kind of VR device and a pocket gadget for the company’s Muse agent. He also used it to answer rivals who want the industry to slow down.
“It’s no secret that people in the industry have a wide range of views about how the future is going to go,” he said, according to the Associated Press. “We are optimists because we believe in people.” That line carried the keynote. Meta is betting that wearable AI, sold to hundreds of millions of people, is a safer path than a small club of labs agreeing to pace themselves.
This article sets out what Meta announced, why Zuckerberg tied his wearable AI plan to the safety debate, what the hardware costs, what the bet costs Meta, and where the argument holds up and where it does not. We covered the privacy side of the glasses the day before the keynote in Meta Leans Into AI and Smart Glasses Amid Privacy Pushback.
Table of contents
- What Meta Announced at Connect 2026
- Why Zuckerberg Framed Wearable AI as an Answer to AI Doom
- Muse Is the Centrepiece of Meta’s Wearable AI Plan
- The Wearable AI Hardware, Device by Device
- What the Wearable AI Bet Costs Meta
- Where Zuckerberg’s Case Against AI Doom Is Strong
- Where the Argument Against AI Doom Is Weak
- What Wearable AI Means for Businesses
- What to Watch Next
- Frequently Asked Questions About Meta’s Wearable AI Push
- References
What Meta Announced at Connect 2026
Connect is Meta’s yearly developer and hardware event. This year almost every announcement served one goal: putting the Muse agent within reach at all times, through something you wear or carry.
The short version
Meta showed new AI glasses, an audio-only pair with no camera, a hearing feature cleared by the US Food and Drug Administration, a VR device the size of glasses, and a keychain-like device called Muse Charm. It also added new features and shopping connectors to Muse itself. The table lists each launch with its price and date.
| Launch | What it is | Price | Availability |
|---|---|---|---|
| Ray-Ban Meta (Gen 3) | Slimmer camera glasses with longer battery life and an action button | From $449 | Available now |
| Ray-Ban Meta Audio | Open-ear audio glasses with AI, no screen and no camera | From $349 | October |
| Hearing enhancement | FDA-cleared, over-the-counter hearing aid feature for supported glasses | $149.99, or with Meta One | US, later in 2026 |
| Meta VR Glasses | About 100 g VR and spatial computing glasses tethered to a compute puck | $1,299.99 | Spring 2027 |
| Muse Charm | Palm-sized device carrying the “whole Muse experience” | Not announced | December holidays |
| Muse updates | Voice mode, avatars, glasses access, an email address and 18 new connectors | Included | Rolling out |
Why the line-up matters
Each wearable AI device is a different way of reaching the same assistant. The glasses give Muse a camera and a microphone at eye level. The audio pair gives it your ears without recording anyone. Muse Charm gives it a voice when you do not want glasses on your face. Meta’s wearable AI plan is less about any single product than about how many places Muse can live.
The keynote’s tone
Zuckerberg called it a “wild time to be building.” The mood was deliberately upbeat. Meta brought a plush-toy mascot for Muse on stage, and senior leaders acted out everyday tasks with it. The message was that personal agents should feel friendly and ordinary, not like the frightening systems at the centre of recent safety warnings.
Why Zuckerberg Framed Wearable AI as an Answer to AI Doom
The most unusual part of the keynote was not a product. It was Zuckerberg’s decision to use a consumer hardware event to take a side in the argument over whether frontier AI development should be deliberately slowed.
The pacing proposal he rejected
Earlier this month, Anthropic chief executive Dario Amodei published a three-step plan that included a proposal for companies and countries to coordinate a slowdown in frontier development. OpenAI’s Sam Altman and Elon Musk at SpaceXAI both backed it. We covered the essay in Anthropic CEO Dario Amodei Calls for Pacing AI Development for Safety.
What he wrote on 15 September
Zuckerberg answered on X a week before Connect. “Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens,” he wrote. He added that AI companies face “significant liability” if their systems cause harm, which he argued is incentive enough.
The Muse delay he cited
He pointed to Meta’s own record as evidence. Meta delayed the release of Muse for several months to make sure it was safe and secure, he said. “We didn’t call for everyone else to do this before we would.” In his view, a lab that slows itself down when it needs to is better than a lab that asks its rivals to slow down with it.
What he said on stage
At Connect he widened the argument. “We believe that the best way to create a positive future for everyone is to put power in people’s hands,” he said. He argued that “the highest purpose of superintelligence is creation and invention — not automation,” and that “the path to a safe future is to distribute this technology widely and create the right checks and balances, rather than concentrating it in the hands of just a few.”
Why wearable AI fits that argument
This is where the hardware and the politics meet. If safety comes from spreading AI widely rather than holding it back, then wearable AI is the purest form of the idea. A device on your face or in your pocket puts an agent in the hands of individuals rather than institutions. Zuckerberg’s case is that this distribution is itself a safeguard.
| Question | Pacing camp (Amodei, Altman, Musk) | Meta’s position (Zuckerberg) |
|---|---|---|
| Who decides the pace? | Labs and governments, in coordination | Each lab, for itself |
| What enforces safety? | Outside evaluators and agreed limits | Liability and each company’s own incentives |
| Where should compute go? | Slower frontier training while safety catches up | “The significant majority” to serving people, not recursive self-improvement |
| Biggest risk named | Systems escaping human control | Power concentrated in “just a few” hands |
| Example offered | Embedded evaluators before release | Meta delaying Muse by several months |
Muse Is the Centrepiece of Meta's Wearable AI Plan
Zuckerberg called Muse the “centerpiece” of Meta’s hopeful outlook. Every wearable AI device announced at Connect is, in practice, a new front door to it.
What Muse is
Muse is Meta’s consumer agent. Unlike a chatbot, an agent can act for you: booking a table, ordering groceries or planning a trip. Meta’s chief AI officer, Alexandr Wang, said on stage that having an agent has been a “sign of someone who is completely AI-pilled.” Meta’s goal, he said, is to make agents accessible and bring “personal superintelligence” to everyone.
New features announced
Meta’s own recap lists a real-time voice mode for long conversations that keeps working on tasks in the background, a customisable voice, and a Muse Realtime Avatar for video chats. Muse now has its own email address. Its Mac app can use your computer to “drive any app on your Mac.” It is reachable through the Muse app, WhatsApp, the web and the Mac.
The connector list
Meta named 18 new connectors: Walmart, Best Buy, American Eagle Outfitters, DICK’S Sporting Goods, Fanatics, Gap, Michael Kors, Sephora, Ulta, Wayfair, Shop Pay, PayPal, Expedia, Instacart, Notion, Granola, GitHub and Box. The AP also named OpenTable, Spotify and Ticketmaster as partners. Most of these are shopping, payment or travel services, so this wearable AI push is also a commerce push.
Muse on the glasses
Meta is bringing Muse to its AI glasses for hands-free use, with visual context. In plain terms, the agent will be able to see what you are looking at and act on it. This is the step that turns wearable AI from a novelty into a platform. It is also the step that raises the sharpest privacy questions.
Muse Charm
Zuckerberg said Meta will release Muse Charm, a keychain-like device with the “whole Muse experience” packed into it, in time for the December holidays. The palm-sized gadget lets people use the agent without a phone, computer or glasses. Meta’s recap says it runs a real-time voice model and that more details will follow later this year. No price was given.
The Wearable AI Hardware, Device by Device
Zuckerberg has long said wearable AI will eventually replace the smartphone as the main way people use technology. Connect was his most complete attempt yet to show what that looks like at different prices.
Ray-Ban Meta Gen 3
The newest version of the classic camera glasses is on sale now from $449. Meta says it is slimmer and more comfortable, with longer battery life and a new action button for quick access to Meta AI. New Aviator and Zena frames join the range.
Ray-Ban Meta Audio
The audio pair is the most interesting product commercially. It plays music, makes calls and talks to the assistant, but has no screen and no camera. The AP noted that video recording “has raised privacy concerns with some users recording people without permission.” Meta’s answer is a wearable AI device that cannot record anyone. It arrives in October from $349, with 12 hours of battery life.
Hearing enhancement
A previous feature helped wearers hear conversations in noisy rooms. Zuckerberg said it now turns supported glasses into an “FDA-cleared, over-the-counter hearing aid” that can “compensate for perceived mild to moderate hearing loss.” It launches in the US later this year for $149.99, or as part of a Meta One subscription, for adults aged 18 and over.
Meta Ray-Ban Display
The glasses with a small built-in display gain cycling and public transit directions in the US and Canada, plus a personalised hologram for video calls. They can now be ordered online, went on sale in Canada and the UK the same day, and open for pre-order in France, Italy and Germany ahead of 13 October availability.
Meta VR Glasses
The biggest wearable AI hardware change replaces the Quest headset with glasses weighing about 100 grams, roughly a fifth of a Meta Quest 3. They tether to a separate puck for battery, storage and compute, use eye tracking and hand gestures instead of controllers, and cost $1,299.99 when they launch in spring 2027.
More frames and more countries
The AP reported more than 100 style options across Ray-Ban, Oakley and Meta glasses by the end of the year, with sales expanding to Turkey, Hong Kong and the Philippines. Meta’s recap adds Singapore, South Korea and Mexico for its Meta Glasses line, with Brazil and the UAE to follow.
The ladder runs from a $149.99 software add-on to a $1,299.99 device, a spread of about 8.7 to 1. Muse Charm has no announced price, so it is not shown.
What the Wearable AI Bet Costs Meta
Optimism is expensive. The hardware behind Meta’s wearable AI push is built by Reality Labs, the division that makes the Quest headsets and, with EssilorLuxottica, the Ray-Ban Meta glasses.
The latest quarter
In its second-quarter results on 29 July, Meta said Reality Labs lost $4.62 billion on revenue of $431 million, according to CNBC. The loss widened from $4.53 billion a year earlier and from $4.03 billion in the first quarter. Revenue rose from $370 million, which Meta put down to strong growth in AI glasses offset by lower Quest sales.
The running total
CNBC reported that Reality Labs has generated more than $80 billion in total operating losses since late 2020. That is the price of Zuckerberg’s view that the next computing platform will be worn rather than held.
The ratio worth noticing
Dividing the quarter’s loss by its revenue gives about $10.70 of operating loss for every $1 of revenue. Wearable AI is growing, but it is growing from a small base. Glasses sales are real, but they do not yet pay for the research that produces them.
Why the shift towards glasses still makes sense
The same results show where Meta sees the payoff. Growth came from AI glasses, not headsets. Meta and EssilorLuxottica sold more than 7 million smart glasses in 2025, as we reported before the keynote. Replacing the Quest brand with Meta VR Glasses reads as an admission that people will wear glasses but will not strap on a headset.
Where Zuckerberg's Case Against AI Doom Is Strong
It is easy to dismiss a wearable AI keynote as marketing. But parts of Zuckerberg’s argument deserve to be taken seriously, especially by anyone who worries about who controls the technology.
Concentration is a real risk
A coordinated slowdown agreed by a handful of leading labs would also entrench those labs. Zuckerberg’s warning about power “concentrated in the hands of just a few” echoes concerns raised by open-source developers and competition regulators alike. Wide distribution of wearable AI is one way to push against that.
He made a specific compute commitment
His X post did not stop at slogans. “Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely,” he wrote. Recursive self-improvement, where AI systems speed up their own development, is the scenario that worries safety researchers most.
Meta did delay its own agent
The claim that Meta held Muse back for several months is consistent with the reporting we tracked before launch in Meta Muse Could Be Announced This Week. A company that absorbs the commercial cost of a delay has at least shown it will act alone.
The audio glasses are a real concession
Launching a wearable AI product with no camera responds directly to the “recording people without permission” complaint. It gives buyers a choice that does not place bystanders under observation.
Where the Argument Against AI Doom Is Weak
The same keynote also exposed the gaps in Meta’s position. Self-regulation only works if the wearable AI product record supports it.
Muse’s own security record
Muse launched with a serious security flaw that had to be patched after release, as we reported in Meta’s Muse AI Assistant Rolled Out With a Serious Security Flaw. A months-long safety delay did not prevent it. That weakens the claim that each lab’s own processes are enough, and it matters more once the agent sits on your face with access to your accounts, your camera and your cybersecurity settings.
Liability is slow
Zuckerberg leaned on “significant liability” as an incentive. But liability works after harm has happened, through courts and settlements that take years. For risks that are hard to reverse, a backward-looking incentive is a weak substitute for checks before release.
Distribution cuts both ways
Putting an agent in hundreds of millions of pockets spreads power, but it also spreads exposure. An agent that can shop, pay and send email for you becomes a target the moment it is widely worn. The wider the wearable AI roll-out, the larger the blast radius of any single flaw.
Partners can still say no
Amazon has already barred Muse from shopping on its site, as we covered in Amazon Bars Meta’s Muse AI From Shopping on Its Site. The new connector list is long, but the biggest retailer is not on it. An agent is only as useful as the services that agree to work with it.
What Wearable AI Means for Businesses
Most organisations will not buy Meta VR Glasses next spring. But wearable AI will reach their premises, their staff and their customers well before that, and it is worth preparing now.
Update your device and visitor policies
Camera glasses, the first mainstream wearable AI category, are already on sale in the UK. Policies written for phones rarely cover a device that looks like ordinary eyewear. Decide where recording is not allowed, such as meeting rooms, clinics and secure areas, and say so plainly.
Treat agents as a new data flow
When staff connect an agent like Muse to email, calendars or work tools such as Notion, GitHub or Box, company data moves through a third party. That needs the same review as any other software supplier. Our AI strategy work starts with mapping exactly these flows.
Plan for agents as customers
If Muse books tables, orders groceries and buys clothes, some of your customers will soon be software acting for people. Retailers should decide whether to support agent traffic, as Meta’s partners have, or block it, as Amazon has. Either choice is legitimate, but it should be a decision rather than an accident.
Watch the accessibility gains
The hearing feature is a reminder that wearable AI is not only about recording and shopping. A $149.99 add-on that works as an over-the-counter hearing aid could matter a great deal to employees and customers with mild hearing loss.
What to Watch Next
Connect set out a timetable. Several dates will test whether Meta’s optimism turns into products people use every day.
October: the audio glasses
Ray-Ban Meta Audio is the first test of whether a camera-free wearable AI device can sell. Its reception will show how much the privacy backlash is costing Meta.
December: Muse Charm
A holiday launch puts Muse Charm, the smallest wearable AI device Meta has shown, up against every other gadget of the season. Price, battery life and what works without a phone will decide whether it avoids the fate of earlier AI gadgets.
Spring 2027: the VR glasses
At $1,299.99, Meta VR Glasses will show whether spatial computing can reach beyond enthusiasts once it stops looking like a headset.
The next safety argument
Zuckerberg has now made his position public twice. If Muse or the glasses cause a serious incident, the case for leaving safety to each lab will be judged on Meta’s own record. We looked at why earlier versions of his AI pitch failed to persuade people in Why People Aren’t Buying Mark Zuckerberg’s AI Future.
Frequently Asked Questions About Meta's Wearable AI Push
What is wearable AI?
Wearable AI means devices worn on the body, such as glasses, earbuds or a small clip-on, that give you access to an AI assistant without using a phone or computer. Meta’s range now includes camera glasses, audio glasses, display glasses, VR glasses and the upcoming Muse Charm.
What did Zuckerberg say about AI doom?
He said Meta is optimistic “because we believe in people” and rejected calls for a coordinated industry slowdown. He argued each lab should set its own safe pace, and that spreading AI widely with checks and balances is safer than concentrating it.
What is Muse Charm?
A keychain-like, palm-sized device that gives access to the full Muse agent without a phone, computer or glasses. Meta plans to release it for the December holidays and has not announced a price.
How much do Meta’s new glasses cost?
Ray-Ban Meta Gen 3 starts at $449 and is available now. Ray-Ban Meta Audio starts at $349 in October. Meta VR Glasses cost $1,299.99 and arrive in spring 2027.
How much money is Reality Labs losing?
It lost $4.62 billion in the second quarter of 2026 on revenue of $431 million, and more than $80 billion in total since late 2020, according to CNBC.
Is Meta’s view on AI safety shared by other labs?
No. Anthropic’s Dario Amodei proposed a coordinated slowdown, and OpenAI’s Sam Altman and Elon Musk backed it. Zuckerberg is the most prominent leader to reject it publicly.
References
Meta bets big on wearable AI, new AI agent as Zuckerberg pushes back against AI doom (AP)
Everything We Announced at Meta Connect 2026
Zuckerberg distances Meta from calls for coordinated slowdown on AI (AP)
Meta’s Reality Labs lost over $4.6 billion in second quarter (CNBC)
Meta Connect 2026 recap (Tom’s Guide)
Meta Connect 2026: New AI and VR glasses and Muse agent get the spotlight
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
