Meta is still internally debating how to handle the privacy LED on its upcoming glasses with always-on AI, The Financial Times reports.
The Information first reported on Meta’s plans to launch smart glasses that could recognize faces and run continuous AI sessions for “hours” over a year ago. The new Financial Times report directly states that it is referring to the same device.
Both reports claim Meta internally refers to the always-on AI feature as “super sensing”. The Financial Times report describes how it would work on a technical level: the system would capture an image from the camera every few seconds and constantly capture audio, and an AI model would continually analyze the images and audio to upload its analysis as text to Meta’s servers.
On a recent Meta earnings call, Mark Zuckerberg told investors that he wants Meta AI on smart glasses to evolve from “being able to answer questions to being able to be a personal agent that’s with you all day long, helping you remember things and achieve your goals”.
According to the Information’s report from last year, the “super sensing” feature could, for example, remind you to pick up your keys if you forgot, or to stop in a store to pick up ingredients for dinner as you walk past it. The report also noted Meta was “exploring” a facial recognition capability, which could for example remind you of someone’s name.
Abrash & Zuckerberg: Future Meta Glasses Will Have Always-On “Contextual AI”
Michael Abrash described how future Meta glasses will have always-on “contextual AI”, and Mark Zuckerberg thinks this will arrive in less than 5 years.
At Meta Connect 2025, the Chief Scientist of the company’s Reality Labs division, XR veteran Michael Abrash, openly described the company’s eventual goal of shipping glasses with always-on “contextual AI” that can even continuously create a dynamic 3D map of your physical environment, and your movements and actions within it, including the objects you interact with. The glasses will store a log of these actions and interactions, and use it to provide “contextual AI”, he said at the time.
For example, you could ask “how many calories have I consumed today?”. And without needing to have logged anything in advance, the AI will be able to answer – as long as you were wearing the glasses at the time.
It’s unclear how wide the gulf between Meta’s first “super sensing” glasses and the eventual device Abrash hopes to ship will be. It seems unlikely, for example, that the first-generation technology will be creating a dynamic 3D map of your physical environment.
Meta Is Improving Its Smart Glasses Privacy LED Tampering Detection
Meta says it’s improving the ability of its smart glasses to detect if the privacy LED has been tampered with and disable the camera, via a mandatory firmware update.
The more pressing issue, according to both reports, is how the privacy issue of “super sensing” would be handled, and how the general public, including bystanders, might react.
One point of internal debate, reportedly, is whether having the feature enabled should continually flash the privacy LED. Meta just announced that it’s improving its detection of buyers tampering with the LED, but the current plan is to not activate the LED during “super sensing”.
The rationale behind this is that the images and audio captured during this mode would not be available to the user, and not even stored after processing. An AI would see almost every moment of your glasses-wearing life, in other words, but no human would – in theory at least.
However, as the report makes clear, a final decision has not been made, and internal debate is still ongoing around how to handle privacy.
In a statement to The Financial Times, Meta said it declines to comment on “internal prototypes” but claimed that it focuses on “privacy built in from the ground up”, and pointed to how its Project Aria Gen 2 research glasses “uses privacy-protective technologies to help people without capturing photos and videos the way traditional cameras work”.
