You’re not the only one seeing what you record on Meta smart glasses, contractors say

Contractors for Meta say they’ve been tasked with reviewing highly sensitive footage recorded by the company’s AI-powered smart glasses. The contractors, based in Nairobi, Kenya, claim to have seen everything from individuals using the bathroom to taking off their clothes.

Speaking with the Swedish newspapers Svenska Dagbladet and Göteborgs-Posten, employees with the data annotation business Sama said that complaints about the review process have been met with swift termination.

“You understand that it is someone’s private life you are looking at, but at the same time you are just expected to carry out the work,” one contractor said. “You are not supposed to question it. If you start asking questions, you are gone.”

Meta has tasked Sama with a job known as data labeling, which requires humans to review and annotate video to train AI models.

The glasses record footage any time a user enables a feature known as “live AI.” The feature allows users to ask questions and receive answers about what they are seeing.

“Your glasses camera and microphone are continuously active during the session so that Meta AI’s responses are faster and related to what you’re seeing,” Meta says.

But what people see and record isn’t private.

Meta’s terms of use state that the company reserves the right to retain and review all interactions with its AI. Meta also says that users should refrain from sharing any sensitive information with its AI, given that it can be reviewed either by automated processes or humans.

But the contractors say Meta has failed to properly inform users that their footage is being seen by humans.

“In some videos you can see someone going to the toilet, or getting undressed,” a contractor said. “I don’t think they know, because if they knew they wouldn’t be recording.”

In one incident, a contractor claims to have clearly seen a user’s debit card, including the account number. 

Other stories are much more graphic. Sometimes the footage captures pornography the users watched. And sometimes the glasses film the users’ own sex acts.

“You think that if they knew about the extent of the data collection, no one would dare to use the glasses,” another contractor said.

Aside from sensitive video, chat logs also reveal intimate details on users’ lives. Another worker reported hearing conversations in which one user discussed criminal activity. Another chat with the AI centered around a user’s attendance at a protest.

Terms of service

The Swedish newspapers say they spent weeks sending inquiries to Meta to no avail. After two months, a Meta representative responded by pointing them to the company’s privacy policy and terms of use.
“When live AI is being used, we process that media according to the Meta AI Terms of Service and Privacy Policy,” the spokesperson said.

Meta did not respond to a request for comment from Straight Arrow News.

Concerns over smart glasses recently led to the creation of a cell phone app aimed at detecting the technology. And although Meta’s glasses include a small LED light that alerts others to when a user is recording, disabling the light has proven easy.

Pushback against the technology has intensified in recent weeks after The New York Times quoted a Meta internal memo from last year saying the company planned to introduce a facial recognition feature when civil liberties groups would be too distracted to protest.