Meta's Superintelligence Labs has rolled out Muse Spark, a generative AI model launching this week through the Meta AI app with plans for broader integration across Facebook, Instagram, and WhatsApp in coming weeks. The company positioned the model as particularly adept at answering health-related questions, stating it collaborated with over 1,000 physicians to refine training data for more accurate and comprehensive responses.
Meta Launches Health-Focused AI Assistant
Testing Muse Spark reveals a concerning feature: the chatbot actively encourages users to upload sensitive health information. When asked about its capabilities, the model explicitly requested raw data from fitness trackers, glucose monitors, and lab reports, offering to calculate trends and flag patterns. This approach mirrors similar features in competing AI tools. OpenAI's ChatGPT and Anthropic's Claude both include health-focused modes allowing users to connect Apple or Android health data directly. Google similarly enables medical data uploads to Fitbit for its AI health coach analysis.
Chatbot Actively Requests Sensitive Health Data
However, privacy experts warn against sharing health data with these systems. Monica Agrawal, an assistant professor at Duke University and cofounder of Layer Health, emphasizes the dual-edged nature of providing detailed health information to AI models. While comprehensive data enables more contextual and personalized responses, it creates substantial privacy vulnerabilities that many users underestimate.
HIPAA Gaps Create Privacy Vulnerabilities
A critical issue stems from HIPAA non-compliance. Unlike medical facilities bound by this landmark U.S. privacy law protecting patient information, mainstream AI chatbots operate under minimal health data protections. Information shared with Meta AI may be retained and used for training future models, according to the company's privacy policy regarding generative AI data retention.
Healthcare Standards Lag Behind AI Practices
Medical professionals express significant concern about this practice. The casual nature of chatbot interactions contrasts sharply with healthcare's traditional privacy standards. Users accustomed to strict confidentiality during doctor visits face substantially looser regulations when uploading clinical results to AI platforms, creating a false sense of security around sensitive personal health data sharing.