ICO writes to Meta over ‘concerning’ AI smart glasses report

The UK’s Information Commissioner’s Office (ICO) has initiated contact with Meta, expressing significant concern following a report detailing allegations that outsourced workers were able to view highly sensitive content captured by the company’s AI-powered smart glasses. The investigation, conducted by Swedish newspapers Svenska Dagbladet (SvD) and Goteborgs-Posten (GP), revealed that subcontracted workers in Kenya may have had access to a range of footage, including intimate moments such as individuals using the toilet or engaging in sexual activity, all recorded through Meta’s Ray-Ban Meta smart glasses.

Meta acknowledged that subcontracted workers might, on occasion, review content captured by its AI smart glasses. The company stated this practice is undertaken "for the purpose of improving the ‘experience’" and is detailed within its privacy policy. However, the Swedish investigation suggests that the filtering mechanisms designed to protect user privacy may not always be effective. One worker reportedly conveyed the unsettling nature of their work, stating, "We see everything – from living rooms to naked bodies."

The Ray-Ban Meta glasses, launched with the promise of integrating AI capabilities for hands-free interaction with the environment, allow users to ask questions and receive answers about their surroundings. Meta’s AI terms of service, specifically within the UK, outline that "In some cases Meta will review your interactions with AIs… and this review may be automated or manual (human)." While Meta asserts that data is "first filtered to protect people’s privacy," sources speaking to SvD and GP claimed that blurring of faces and other identifying features sometimes failed, leaving individuals’ identities exposed.

Users are required to manually activate recording or use a voice command, but the report highlights a potential lack of awareness that their videos and images could be reviewed by human operators. Although Meta’s extensive privacy policies and terms of service do mention the possibility of manual review, the explicit nature of the content allegedly viewed raises substantial privacy questions.

The ICO, the UK’s data protection regulator, has responded with a strong statement, emphasizing that "devices processing personal data, including smart glasses, should put users in control and provide for appropriate transparency." The watchdog underscored the obligation for service providers to "clearly explain what data is collected and how it is used." In light of the disturbing claims, the ICO confirmed, "We will be writing to Meta to request information on how it is meeting its obligations under UK data protection law."

The workers involved in the review process are reportedly data annotators, tasked with teaching Meta’s AI to interpret images and sounds by manually labeling the content. They also review transcripts of AI interactions to assess the accuracy and adequacy of the AI’s responses. While workplace privacy measures were reportedly in place, including surveillance cameras and restrictions on mobile phone use, the content encountered was frequently described as highly sensitive. Reports included instances of users watching pornography and, in one particularly concerning case cited by the Swedish press, a pair of glasses left recording in a bedroom inadvertently captured a woman undressing.

Meta’s smart glasses feature a visible light indicator that illuminates when the camera is actively recording images or videos. The company’s guidelines advise users to be considerate of others by indicating when recording is active and to refrain from recording in private spaces.

The broader context of this report involves the rapid advancements in artificial intelligence and the subsequent proliferation of wearable AI-enabled gadgets. These devices offer a range of functionalities, from real-time text translation to providing descriptive information about the user’s visual environment, which can be particularly beneficial for individuals with visual impairments. However, this growing adoption has been accompanied by escalating concerns regarding the potential for misuse. Previously, women have shared accounts with the BBC of being filmed without their consent by individuals using smart glasses.

The data annotators mentioned in the report work for Sama, a company that originated as a non-profit organization with the stated mission of creating employment opportunities through technology. While Sama holds an "ethical" B-corp designation, it has faced scrutiny and legal challenges in the past related to its content moderation services for other tech companies. Following criticism and legal action from former employees, Sama ceased its content moderation services, with the company later expressing regret for having undertaken such work. The BBC has reached out to Sama for comment on the recent report.

Meta’s parent company, EssilorLuxottica, the manufacturer of the Ray-Ban Meta glasses, has also been approached by the BBC for comment. The implications of this report extend beyond Meta, highlighting the complex ethical and privacy challenges inherent in the development and deployment of AI-powered consumer technology. As these devices become more integrated into daily life, the onus is on both manufacturers and regulators to ensure robust data protection measures and clear communication with users about how their information is being collected and utilized. The ICO’s intervention signifies a critical step in holding tech giants accountable for the privacy implications of their innovative, yet potentially intrusive, technologies. The potential for sensitive personal data to be accessed by third-party contractors, even with purported filtering and privacy safeguards, raises fundamental questions about user consent, data security, and the responsibility of companies in the AI era. The public’s trust in these emerging technologies hinges on transparency and a demonstrable commitment to protecting individual privacy.