Privacy Alert: Investigation Suggests That Meta Smart Glasses Record Users Without Consent

A recent investigation by Swedish news outlets Svenska Dagbladet and Göteborgs-Posten has revealed significant privacy breaches involving Meta’s smart glasses. Findings indicate that “AI annotators” based in Nairobi, Kenya, have been reviewing private imagery captured by the devices to train Meta AI. This data includes sensitive content such as domestic interiors (kitchens and bathrooms), nudity, sexual acts, and personal financial documents.

Key Findings and System Failures

The investigation highlights several critical issues regarding how Meta handles user data:

• Identification Risks: While Meta utilizes automated face-blurring, the technology reportedly fails frequently, leaving users identifiable.

• Policy Shifts: Privacy concerns escalated following a 2025 policy update. This change allows cameras to remain active whenever the “Hey Meta” voice command is enabled and removes the ability for users to opt out of cloud storage for voice commands.

• Unintended Capture: Despite marketing the glasses as a style assistant for choosing outfits, the always-on potential of the hardware leads to the recording of private moments, such as dressing or using the restroom.

Legal and Regulatory Response

The disclosure has triggered significant backlash from privacy advocates and legal bodies:

• Class Action Lawsuit: A suit filed in California alleges that Meta engaged in deceptive advertising by claiming the product was “privacy-first” while allowing third-party contractors to view intimate footage.

• Regulatory Scrutiny: The UK Information Commissioner’s Office has questioned Meta regarding these allegations, while the Electronic Privacy Information Center (EPIC) labeled the device a “serious risk to civil liberties.”

Meta’s Position

Meta spokesperson Tracy Clayton stated that media remains on-device unless users choose to share it with Meta AI. The company maintains that using human contractors to review shared data is a standard industry practice aimed at improving AI performance, asserting that filtering measures are in place to protect user identity.