Meta Might Be Using Photos Taken on Ray-Ban Smart Glasses to Train AI
When asked this week if Meta is using photos taken on Ray-Ban smart glasses to train its AI, company representatives refused to deny or confirm.
In an interview with Tech Crunch on Monday, Meta was asked if it plans to train its AI models on the images taken by customers’ Ray-Ban Meta Smart Glasses the same way it uses photos from public Instagram and Facebook accounts.
“We’re not publicly discussing that,” Anuj Kumar, a senior director working on AI wearables at Meta, tells Tech Crunch in a video interview.
“That’s not something we typically share externally,” adds Meta spokesperson Mimi Huggins. When Tech Crunch pushed for further clarification Hussing replied, “We’re not saying either way.”
For those not familiar with Ray-Ban Meta smart glasses, they can shoot both 1080 30P video and 12-megapixel stills. Furthermore, a real-time video update is coming to the glasses which will stream images into a multimodal AI model so the glasses can answer questions about what the user is looking at.
This will mean that the smart glasses will be gathering even more images than before and as Tech Crunch notes, the user might not be completely aware the device is recording imagery. In private moments, the user might be asking the glasses about an object they’re looking at — unaware the images are being saved in a cloud.
The obvious problem with smart glasses has always been the fact that they could be used as a clandestine recording device. In his review of the Meta Ray-Bans, PetaPixel’s Chris Niccolls wrote.
“There is an LED light that flashes whilst recording stills and video. To be fair, it is quite bright but I still feel that most people won’t notice. If the LED is obscured the glasses will not record but a video can be started and will continue to record if the LED is covered afterwards. I’ll leave it up to you the reader to determine how and when you feel comfortable shooting.”
While Meta and many other AI companies argue that imagery on the open web and on social media is “publicly available”, it is far more difficult to argue that imagery taken on a private person’s smart glasses is public.
The fact that Meta won’t say that it isn’t using imagery from smart glasses for AI training purposes is alarming given the kind of private imagery the devices are capable of collecting.
Update 10/3: After publication, a Meta spokesperson confirmed that it may train its AI on any image a user asks Ray-Ban Meta smart glasses to analyze. “In locations where multimodal AI is available (currently U.S. and Canada), images and videos shared with Meta AI may be used to improve it per our Privacy Policy,” Meta policy communications manager Emil Vazquez tells Tech Crunch.