Note: Webitlio has a partnership with Proton, whose analysis this article draws on. The article contains no affiliate links.
Smart glasses look like a simple gadget: a camera, speakers and an AI assistant that can answer questions about what you see. A Proton analysis argues they are also something else, a source of the first-person view of daily life that AI systems for robots lack. The piece is an opinion, and the link between Meta’s glasses and its robots is a hypothesis, not a documented fact. The pieces it points to are real, though, and worth laying out.
Why robots need first-person data
Language models learn from text, but a robot has to deal with the physical world: where objects are, what happens when they move, and how hands handle things that bend, spill or break. Proton cites an argument by Meta’s former chief AI scientist Yann LeCun that a four-year-old has already taken in far more sensory data than the largest language models, which is why machines need richer input.
Researchers are working on several problems that first-person video can help with. One is episodic memory, such as remembering where you left your keys. Another is intuitive physics, such as predicting what happens when a ball rolls toward a table. The third is hand-object interaction, such as tying laces or handling cloth. A camera worn on the face sees the world from the same viewpoint as the person doing the task, which is what makes this footage valuable.
What Meta says its glasses capture
According to Proton’s summary of Meta’s position, Ray-Ban Meta glasses do not continuously record everything for AI training. The data Meta uses includes voice commands and the video captured when a user asks the AI to do something. When the assistant needs to analyze what the camera sees, the footage can be sent to Meta’s servers. Live AI removes the usual wake phrase, which means the camera and microphone can stay active for longer.
Meta added a new control on September 23. Meta says that if you opt out and ask your glasses to translate a menu, it uses your visual data to answer and then deletes it, and the data is not used for training or reviewed. As reported by iDevice, the setting covers only what the camera sees. Voice requests can still be used for training and review, although you can delete those recordings yourself. Meta also plans to bring its encrypted Private Processing setup to the glasses, but an engineering director said there is no timeline.
The bystander problem
The wearer controls the settings, but the people around them do not. In Proton’s 72-hour test, faces were blurred in the material that was reviewed, but the people nearby did not know they were being recorded. Meta’s research program, Project Aria, works differently: its published guidelines say participants are trained in recording practices, people in private homes must consent, and a visible light shows when audio and video are being collected. Consumer glasses on a street or in a shop have none of that structure.
What points to Meta’s robotics plans
Proton lists the facts that make the hypothesis plausible. Meta created a humanoid robotics division inside Reality Labs, the organization behind the glasses, in February 2025. On May 1, 2026, it acquired Assured Robot Intelligence, a startup whose team joined Meta Superintelligence Labs and works with Meta Robotics Studio on humanoid hardware and AI, according to The Robotics Media. Reports also say Meta wants to supply the software and tools for other companies’ humanoid robots. The startup’s co-founders have said that learning from human experience, not teleoperation alone, is how they expect robots to improve.
What is not established
None of this shows that Meta is using glasses footage to train robots. Proton says so itself: the evidence explains why such data could be strategically valuable, not that it is being collected for that purpose. We found no Meta statement saying the glasses feed its robotics work, so treat the connection as a question for Meta to answer.
What you can do
If you own the glasses, you can open the Meta AI app, go to your glasses settings, choose Meta AI and turn off “Store visual data from AI experiences,” according to Let’s Data Science. Opting out may disable features that rely on stored visual data, such as visual reminders. Voice recordings must be deleted by hand. Bystanders have no setting at all, so the practical rule is to keep AI camera features off in places where other people expect privacy.
The bigger question
As AI moves from the screen into the physical world, first-person data becomes more valuable. The useful questions are the ones Proton ends on: what a device can see, hear, store and send away, and who decides.
