Offshore Meta Contractors Reportedly Reviewing Footage of Users in Intimate Settings; Meta has Failed to Explain How Sensitive Images Are Used to Train AI Systems
Washington, D.C. – U.S. Senator Ron Wyden, D-Ore., and Senator Edward J. Markey, D-Mass., today launched an inquiry into reports of severe privacy violations enabled by Meta’s AI-enabled smart glasses, including capturing sensitive imagery of users and bystanders without their consent, sending intimate images to overseas contractors, and using highly sensitive images to train Meta’s AI systems.
“Although Meta identified risks to bystander privacy and vulnerable populations before launching its glasses in 2021, it nevertheless launched a product tailor-made for invading the privacy of its users and bystanders,” Wyden and Markey wrote to Meta CEO Mark Zuckerberg. “Meta’s professed commitment to privacy and preventing intimate image abuse rings hollow if the company cannot demonstrate meaningful steps to prevent its own products and practices from facilitating those harms.”
Meta smart glasses have been linked to several invasions of users’ and bystanders’ privacy since the product’s launch in 2021, despite the company identifying risks to bystander privacy and vulnerable populations in advance of the product’s release. News outlets have cited numerous reports of the use of Meta glasses to sexually harass unsuspecting women for social media content. For example, in October of 2025, 404 Media reported that a network of men secretly filmed women while soliciting sexual services using Meta AI Glasses. The recordings were distributed onto social media platforms that garnered millions of views.
The letter follows a request in March, led by Markey, seeking information about Meta’s reported plans to integrate facial recognition technology into Meta Glasses.
In the letter, Wyden and Markey requested answers to the following questions by no later than November 25, 2026.
- Meta’s 2021 Human Rights Report states that the company conducted an internal human rights analysis before launching Ray-Ban Stories, identifying risks to bystander consent and privacy, particularly for women, children, and other vulnerable populations. Please provide that analysis and all subsequent human rights impact assessments, privacy and safety assessments, and related due diligence concerning Meta’s camera-equipped glasses from initial development to the present.
- In conducting its human rights due diligence, did Meta specifically evaluate the potential for its glasses to facilitate voyeurism, nonconsensual intimate image capture, or similar privacy invasions, distinguishing by risks arising from initial capture, local storage, transmission to Meta, access by employees or contractors, use in AI training, and subsequent disclosure or distribution? For each risk identified, please describe the safeguards considered and implemented, their effectiveness, any residual risk accepted before product launch or subsequent deployment, and Meta’s justification for proceeding despite that residual risk.
- What testing has Meta conducted of accidental camera activation, recording-indicator tampering, and camera-based AI interactions that do not activate the outward-facing recording indicator? What testing has Meta done to understand the circumstances in which Meta glasses’ recording indicator may not be visible to bystanders, either due to ambient conditions, indicator interference, or for bystanders who are low vision or blind?
- Meta’s documentation indicates that it automatically blurs faces and license plates in certain visual AI interactions, and it employs other technologies to identify and prevent the further distribution of nonconsensual intimate imagery on Facebook and Instagram. What comparable safeguards, if any, does Meta employ in its glasses to prevent the nonconsensual capture, disclosure, or processing of sensitive and intimate imagery?
- Please describe the circumstances under which imagery captured or processed through Meta’s glasses are made accessible to employees or contractors, or used to develop, evaluate, or train AI models, and the access restrictions and filtering procedures that apply during Meta’s processing of that data. Please identify all contractors and subcontractors currently authorized to access such data, the countries in which review occurs, and copies of the contractual provisions governing their access, use, retention, and disclosure of this data.
- Does Meta use all imagery eligible for AI training, or does it select a subset? If the latter, what criteria govern that selection? What safeguards, if any, does Meta have in place to prevent its own use, disclosure, publishing, and processing of intimate imagery to which the subject depicted never consented?
- Meta has publicly emphasized the importance of preventing nonconsensual intimate image abuse. To what extent did Meta’s teams responsible for women’s safety and mitigating intimate image abuse participate in the design, development, and evaluation of Meta’s glasses?
- Can individuals—both device owners and people whose images the glasses capture—request deletion of biometric and other data captured by its glasses? If so, how does Meta ensure timely and complete deletion?
- What other redress mechanisms, if any, are available to users and bystanders who discover that intimate or other sensitive imagery was captured by Meta’s glasses and subsequently disclosed to the public or utilized for AI training without their consent?
The text of the letter to Meta is here.
###