Meta reportedly testing prototype AI specs that record everything the user sees and hears

Meta Platforms Inc. this week sought to reassure consumers about the privacy safeguards of its controversial artificial intelligence-enabled glasses, yet at the same time it’s reportedly pushing even creepier capabilities in its “internal prototypes.”

The company’s AI-powered eyewear has a growing reputation as a creepy technology, but in a blog post Tuesday the company announced it’s taking steps to change that opinion. It revealed it’s making a key update that will disable its glasses’ cameras if the LED light that’s meant to indicate they’re recording is blocked or tampered with.

The move appears to be a concession to the growing sentiment that AI specs aren’t just fashionable high-tech accessories, but also have serious implications with regards to people’s privacy. It has become quite apparent that the specs can be abused in nefarious ways, such as by secretly recording people without their knowing.

That’s why it’s concerning that Meta is now testing a new prototype of its smart glasses that will record everything, all of the time. That’s according to a report by the Financial Times, which said the new “super-sensing” glasses would take photos continuously, every few seconds, in order to record everything that the wearer has seen and done throughout the day. The idea, according to sources with knowledge of the project, is that the AI glasses could act as an “always-on assistant” and help to summarize people’s days, tell them where they left their car keys or something that someone told them earlier in the day.

The most worrying aspect of this is that senior Meta executives aren’t planning to activate the LED light when these super-sensing features are being used. If so, it would be much harder for bystanders to know when they’re being recorded.

That’s a significant change from how Meta’s existing AI glasses work. With current models, users have to activate the specs to record their surroundings by using their voice to tell them to do so, or by touching a physical button on the glasses. When they do this, the LED indicator at the front of the glasses lights up to show everyone else that the camera is switched on.

Meta seems to think it can justify not switching the LED light on. In a policy paper published last year, it said that if the LED light were blinking all of the time, people would likely stop noticing it, leading to reduced awareness of when photos or video was being captured by users.

To ease privacy concerns further, the plan is that the raw footage recorded by the super-sensing specs would not be stored by the company, nor would it be made available to the user. Rather, it would simply extract and store the metadata of those images and the audio it records, so it can be queried by AI models. The thinking is that this would have fewer privacy implications, the Financial Times said. However, the metadata can still contain information such as exact locations, device serial numbers and so on.

In a statement to the Financial Times, Meta said that its approach towards its AI glasses is always focused on “privacy built in from the ground up.”

“While we don’t comment on internal prototypes, we’re committed to getting our glasses right because they need to be loved by both the people wearing them and those around them,” the company said.

Meanwhile, there are divergent views within Meta on whether the data collected through the super-sensing feature should be used for AI training purposes. At present, the company already stores and collects data from existing specs users, and that practice has been a cause for concern. A report by the Swedish newspaper Svenska Dagbladet earlier this year revealed that some photos and videos recorded by AI glasses wearers were sent to human contractors based in Kenya to be reviewed and labeled for AI training.

But those videos contained footage of people having sex, using the bathroom and looking at sensitive data such as financial transaction records and passwords. Given that these moments were all captured by devices that have to be switched on manually, it’s almost inevitable that an always-on device would sweep up all kinds of sensitive footage.

The one saving grace is that it’s debatable if Meta can actually bring its super-sensing glasses to fruition. Though founder Mark Zuckerberg might seem to like the idea, it’s not clear if such glasses would have the necessary battery life to be able to make the feature possible. Recording continuously throughout the day would rapidly drain the device’s power. Workarounds might be possible, such as an extremely low-resolution camera that uses less energy, or some kind of bigger battery or even an external one, but it’s not clear how practical those solutions would be.

Photo: Meta Platforms