Meta's AI Glasses Face 70-Plaintiff Federal Lawsuit Over Intimate Footage Routed to Kenya Contractors
A consolidated federal class action filed in Northern California accuses Meta Platforms of building a mass surveillance pipeline inside its Ray-Ban smart glasses, routing footage from intimate settings to low-paid contractors in Kenya without users’ knowledge or consent.
The complaint, amended August 31 in Case No. 3:26-cv-01897-EMC before Judge Edward Chen, runs 230 pages and covers 70-plus named plaintiffs. The class is proposed, not yet certified. Plaintiffs include people who bought the glasses, people who merely wore them, and bystanders who were recorded but never purchased or configured a device.
What the lawsuit alleges
The complaint’s central claim: every activation of Meta AI on the glasses, intentional or accidental, triggers a pipeline that transmits captured video and audio to Meta’s servers, routes the content overseas, and puts it in front of human workers who watch, label, and embed it into Meta’s training data.
The subcontractor is Sama, based in Nairobi. Workers labeling footage for Meta reported seeing people showering, having sex, changing diapers, and discussing medical and financial details. A February investigation by Swedish newspapers Svenska Dagbladet and Göteborg-Posten, based on interviews with more than 30 Sama employees and former Meta staff, found that faces — which Meta says are blurred before review — sometimes appeared unblurred.
Activation can happen without a deliberate “Hey Meta” command. The June 2026 generation of glasses added an “action button” on the inner right temple arm. Plaintiffs allege the button activates AI when glasses are adjusted or when the button incidentally contacts the wearer’s temple.
One California plaintiff, referred to as PL18, alleges he left the glasses on a bathroom counter, unaware they were still active. Photos of family members using the bathroom and bathing subsequently appeared in his gallery.
Sales growth complicates Meta’s position
Meta sold approximately 2 million glasses across 2023 and 2024 combined. In 2025 alone, it sold 7 million. The installed base makes the allegation that intimate footage reached external contractors a large-scale privacy event rather than an edge case, if proven.
The company has responded to the lawsuit by defending its standard data practices. “If you use Meta AI, we may review that data to help improve our products and people’s experiences — this works the same way as many other companies. We take steps to filter this data to help remove identifying information and to protect people’s privacy.” Meta said it disagrees with the allegations and intends to fight them.
The Illinois facial recognition case
A separate lawsuit filed in September in Illinois federal court alleges Meta has used public Facebook and Instagram profile images to train the AI powering NameTag and related tools. Meta CTO Andrew Bosworth discussed NameTag’s capabilities in a July podcast. The California complaint also asks the court to block deployment of biometric identification through the glasses.
Critic’s video pulled from Meta platforms
On September 24, a video critical of the glasses was removed from Instagram under Meta’s bullying and harassment policy. The video, filmed in the Netherlands by Dutch creator Roel Maalderink, used Meta’s own glasses to document an encounter with Meta employees in a public space. The video remained viewable on YouTube. The removal drew attention on Hacker News, with commenters noting that Meta has declined to enforce similar policies against pickup-artist content filmed on the same glasses that has spread across its own platforms.
The removal coincides with ongoing questions about Meta’s tamper-detection claims. Meta announced on September 2 that it had updated the glasses to disable recording if the capture LED is covered, a response to an August CBS News test that showed gaffer tape over the LED did not stop recording on earlier firmware.
Stakes
Meta presents the glasses as central infrastructure for its AI strategy. The complaint describes first-person video as “some of the scarcest and most expensive inputs in generative AI” with an established market price. If the plaintiff class is certified and the footage-routing allegations are proven, the liability exposure extends to millions of activations across a 7-million-unit installed base.
The case proceeds in federal court. No trial date has been set.