Apple leaks camera AirPods built for Visual Intelligence

Share
Apple leaks camera AirPods built for Visual Intelligence

A demo video buried in Apple's latest macOS release candidate shows the next AirPods watching the room — not to shoot photos, but to give Siri eyes.

Apple appears to have accidentally shipped a launch clip for camera-equipped AirPods inside the macOS Tahoe 26.7 release candidate, and the pitch is Visual Intelligence as something you wear, not something you point a phone at. MacRumors found the video. It shows a man in the new buds holding up a book so the camera can read the cover, over a Siri voiceover: "With Visual Intelligence, your world becomes savable. See something you like? Just ask me to save it for later." The software also warns if hair covers the cameras: "To get the most accurate information about things in your environment, make sure AirPods are not covered."

That is the product in one sentence. Tiny cameras on the stems, aimed at whatever the wearer is looking at, feeding Apple's vision stack so Siri can name objects, save them, and answer questions about the physical world. Mark Gurman at Bloomberg has been reporting this hardware for months under the B790 codename: low-resolution visual intake, stems longer than AirPods Pro 3 to house the sensors, an LED that lights when data goes to the cloud, and a possible September debut next to the rebuilt Siri and the next iPhones. The leaked clip only shows the buds from behind, so the indicator light is not visible — but the thicker Pro-like silhouette matches what he described.

The competitive frame is obvious. Meta already sells camera glasses. Looki and Rokid sell pendants and frames that do a version of the same trick. Apple is reportedly building glasses of its own. Camera AirPods are the bridge: a form factor hundreds of millions of people already wear, discreet enough that upgrading existing customers is easier than talking them into a new face computer. Prior reporting says the cameras are not designed to take photos or video. That distinction will not survive first contact with a locker room.

Privacy is the actual story. Meta's Ray-Bans at least look like glasses with a camera on them. AirPods sit in your ears. Bystanders will not know they are being seen. Germany already moved to treat Meta's frames as a recording device disguised as an everyday object — we covered that fight in Attack exposes hidden reasoning traces inside Claude, GPT, Gemini. Apple's version is more discreet, not less. An LED that signals cloud upload is notice for the wearer, not for everyone else in the cafe. The usual Apple privacy pitch — on-device processing, Private Cloud Compute — answers the wrong question. The wearer is not the person who did not consent.

If a demo is already sitting in a release candidate, September is a live date. That would put camera AirPods on shelves in the same season as the new Siri, which is the only reason this hardware exists. Apple does not need another pair of earbuds. It needs a reason for people to talk to Siri while they walk around the world.

What to watch: whether Apple ships an unmistakable recording indicator for bystanders, or bets that "they're sensors, not cameras" will hold.

Would you wear earbuds that can see the room if the people around you never get a vote? Tell us in the comments.

Read more

Brain organoids classified Braille at 83% — in a trio, not alone

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base. The run itself is the number worth keeping: 16,000 Nvidia GB200 GPUs for 38 days across August and September, at 21% average model

Amazon's new Bee devices listen all day — for $11.99 a month

Amazon's new Bee devices listen all day — for $11.99 a month

Amazon turned its always-listening Bee line into a two-device family on Thursday, and NVIDIA opened a 2.8-trillion-token Indic corpus recipe. The connecting thread: both are bets that the next AI money is in memory — yours, or your model's. Amazon put two new always-listening devices on sale, and both now lead to a subscription. Bee Frontier, the next-generation wristband that also clips to clothing, ships now at $74.99 in Onyx, Lavender and Sage — up from the $50 predecessor — with seven-day

Perplexity open-sources its agent sandbox after 118 failed escapes

Perplexity open-sources its agent sandbox after 118 failed escapes

Perplexity handed its agent containment layer to the open-source world, and Washington's AI task force stopped reading briefings and started working. Perplexity open-sourced SPACE on Thursday — the sandbox it built to keep its own AI agents from getting out — and says nine frontier models failed to break out of it across 118 attempts. Each agent runs inside a small virtual machine with its own operating system, with credentials and network access managed outside the sandbox so an agent can re