Leaked docs reveal Meta’s next-gen, 'always on' AI glasses could record without warning light

Meta is reportedly developing next-generation AI glasses that can remember what users see and hear throughout the day – but plans to potentially disable the recording warning light have reignited concerns over privacy, consent and covert surveillance.

The leaked plans surfaced in the Financial Times on Wednesday, and suggest that Meta’s wearable capture is quietly moving from record-on-command to record-by-default.

According to the FT, which cited multiple people familiar with the project, Meta is testing a new "always aware” feature internally known as “super sensing.”

The AI would capture the audio and periodic images throughout the day, allowing users to ask questions such as where they left their keys, who they met, or what was discussed in an earlier conversation.

One proposal floated was that the raw audio and images would not be permanently stored or available to users.

Instead, the system would extract metadata (information about what was seen or heard) so the AI can answer questions while reducing storage of the original recordings. Meta is also considering whether such data could help train its AI models.

Lights out?

Another proposal around the new model was whether or not the recording light should be switched on while the background AI feature is actively recording.

Right now, the white LED light on Meta’s Ray-Ban and Oakley smart glasses is designed to alert people nearby, whenever photos or videos are being captured – but there is currently no signal to indicate when AI mode has been switched on.

Meta has previously described the light as an important privacy safeguard, even announcing this week that existing glasses will disable the camera if the indicator light is physically blocked or tampered with.

In a FAQ blog posted on Tuesday to answer questions on the glasses, Meta said:

“Beginning with our second generation of glasses, the camera is automatically disabled if we detect that the capture LED has been blocked. No photos or videos can be taken until we detect that the light is unblocked.

“We’ve seen some people go beyond using tape to sophisticated efforts to modify or destroy the capture LED.

“We’ve seen some people go beyond using tape to sophisticated efforts to modify or destroy the capture LED."

Meta glasses FAQ page

We are continuously improving our ability to detect tampering, and now we’re updating the glasses to disable the camera if they detect the LED was physically tampered with or destroyed.”

Software update

According to the FT, some of the glasses’ new features could arrive through a software update, rather than requiring users to have to buy a new model of glasses. The internal names for the prototypes sounds like cocktail hour at Meta: The sunglasses model has been named Aperol and the prescription/optical model Bellini.

Meta declined to comment on internal prototypes, telling the FT that it does not discuss projects under development.

While the reported plans reflect CEO Mark Zuckerberg’s ambition to make AI glasses the next major computing platform, that vision comes with privacy trade-offs, with critics arguing that the ability to help someone find their keys should not trump the privacy and safety of women and children.

Despite endorsements by celebrities such as Kylie Jenner, pushing Meta’s AI glasses into the mainstream, the devices have been dubbed “pervert glasses” by some critics and their wearers “glassholes” following reports of wearers filming women without their knowledge.

In its article on ‘The rise of the Meta Ray Ban Creep’ Wired reports on influencers using glasses to record themselves approaching women in public, raising concerns that wearable AI is helping to normalize recording people without asking their permission first.

Curious what others think about this story? Contribute your thoughts to the debate below.

It’s not the first time Meta’s glasses plans have come under scrutiny in recent months. In June two security researchers found that Meta has built a complete facial recognition system into its smart-glasses app, which could be pushed out via an update, although once the story was widely covered, the tech giant reportedly deleted the code.

This week it was announced that eyewear and headwear that contains recording technology will be banned in New York State courts.