An extreme close-up of an eye laced with a red neural web, representing constant recording and surveillance through camera glasses.
Face-worn cameras turn ordinary eye contact into a potential recording, which is exactly what regulators and plaintiffs now question.

Meta’s AI camera glasses, built with eyewear maker Luxottica, are under scrutiny on several fronts in 2026. On 4 March 2026, plaintiffs Gina Bartone (New Jersey) and Mateo Canu (California) filed a class action against Meta Platforms and Luxottica of America. The complaint alleges Meta paired privacy-centric marketing, including the claim “designed for privacy, controlled by you”, with insufficiently clear disclosure about transmission, cloud processing, and human review of captured media.

The disclosure question grew sharper after a Swedish investigation, reported by newspapers including Svenska Dagbladet and Göteborgs-Posten, found that workers at a Kenya-based subcontractor reviewed footage from customers’ glasses. That footage included sensitive content such as nudity, people having sex, and using the toilet. Meta said it blurs faces in images before review, but sources disputed that the blurring worked consistently.

The concerns are not limited to what happens after capture. A BBC investigation found men using smart glasses to covertly film women who did not know they were being recorded. Because the camera sits on the face and blends into ordinary interaction, recording becomes ambient and hard to detect. The UK regulator, the Information Commissioner’s Office (ICO), has begun investigating. Separately, security researchers reported by Malwarebytes in June 2026 found unreleased face-recognition code inside the Meta AI companion app that powers the glasses, raising the prospect of identifying strangers at a distance without consent.

Step 1 Capture A face-worn camera records video and images during ordinary interaction, often without a clear signal to bystanders.
→
Step 2 Transmit Media moves to cloud processing. The complaint alleges buyers were not clearly told this happens.
→
Step 3 Human review Subcontracted workers, reported in Kenya, review footage to label data, including intimate content.
→
Step 4 Scrutiny A class action, an ICO investigation, and researcher findings follow the disclosures.

The central tension is between marketing and consent. Meta promoted the glasses as privacy-first. Plaintiffs argue that ordinary buyers never understood what actually happens to captured data, even where a privacy policy technically mentioned human review. The complaint frames that gap as false advertising and a consumer-protection problem, not a mere technicality.

Why it matters

Always-available, face-worn cameras change the social norms that usually govern filming. A phone raised to record is visible, and people around it can object or move away. Glasses collapse the distinction between social interaction and data capture, which makes recording frictionless and difficult to perceive. That shift affects bystanders who never bought the product and never agreed to anything.

The Kenya footage review reframes what “AI processing” means to a buyer. Many people assume a model handles their media automatically. In practice, human reviewers may watch intimate moments to label training data. When face blurring does not work consistently, the promise of anonymised review weakens. The unreleased face-recognition code deepens the worry, because identifying strangers in real time without consent is a step beyond passive recording.

Regulators are responding on multiple tracks. The ICO investigation signals that UK data-protection law applies to wearable capture, not only to platforms. In Europe, systems that enable this kind of monitoring intersect with the EU AI Act risk framework , which treats certain biometric and surveillance uses as high risk. For builders, these cases are a concrete lesson in AI safety and honest disclosure: what you say in marketing must match what your data pipeline actually does.

Further reading

Sources