Meta AI glasses leak intimate videos to human moderators, sparking privacy outcry

Meta’s Ray‑Ban Display glasses and other AI‑enabled wearables now appear to be funneling raw visual streams to a human‑in‑the‑loop pipeline that operates largely out of Nairobi, Kenya. According to Engadget, which is drawing on a Swedish Svenska Dagbladet investigation, moderators in Kenya have reported seeing “people nude, using the toilet and engaging in sexual activity, along with credit card numbers and other sensitive information” while annotating footage captured by the devices (Engadget, March 3 2026). The company’s terms of service explicitly permit such human review, stating that “either humans or automated systems may review sensitive data” and placing the onus on users to avoid sharing private material (Svenska Dagbladet). Meta’s own response was limited to a boilerplate comment that live AI content is processed under its AI Terms of Service and Privacy Policy, without addressing the geographic scope of the review process.

The practice hinges on Meta’s need to train its large language models (LLMs) to understand visual context. Engadget notes that the “AI ‘annotation’” workflow requires human workers to label what the AI sees, a step Meta says is essential for model improvement. However, the same report highlights that the workers are “underpaid” and that the data they handle can include “intimate video and sensitive financial information” from European users, raising questions about compliance with the EU’s General Data Protection Regulation (GDPR). A data‑protection lawyer quoted in the Svenska Dagbladet piece emphasizes that GDPR mandates transparency about how personal data is processed, a requirement that appears to be at odds with Meta’s opaque privacy policy for its wearables (Svenska Dagbladet).

The privacy implications extend beyond the immediate breach of user expectations. GDPR not only obliges companies to disclose processing activities but also to ensure that data transfers outside the European Economic Area meet strict safeguards. If Meta is routing European‑origin footage to Kenyan moderators without explicit, informed consent, it could be violating the regulation’s cross‑border data‑flow rules. The Swedish newspaper reported that journalists “had to jump through some hoops” to locate the relevant privacy policy, suggesting that Meta’s disclosures are buried and difficult for consumers to find (Svenska Dagbladet). Such opacity fuels the current outcry, as privacy advocates argue that users are effectively forced to trade intimate moments for the convenience of on‑device AI assistance.

Meta has not offered a detailed rebuttal, but the incident underscores a broader industry tension between rapid AI development and responsible data stewardship. The company’s reliance on low‑cost annotation labor mirrors practices seen across the AI sector, where massive datasets are often curated by outsourced workforces. Yet the public backlash against Meta’s glasses may pressure the firm to redesign its data pipeline, perhaps by increasing on‑device processing or by providing clearer opt‑out mechanisms for sensitive content. Until then, European regulators are likely to scrutinize the case, and consumer‑rights groups are expected to file complaints alleging GDPR violations.

The episode arrives at a moment when wearable AI is gaining traction, with competitors promising similar “always‑on” visual assistants. If Meta’s approach proves untenable, it could set a precedent that forces the entire market to confront the hidden human labor behind ostensibly private devices. For now, users of Meta’s smart glasses in Europe face a stark choice: continue to enjoy AI‑enhanced vision while their most private moments may be reviewed by strangers half a world away, or abandon the technology until clearer safeguards are put in place.