Meta’s New AI Asked for My Raw Health Data—and Gave Me Terrible Advice
Meta recently launched Muse Spark, its first generative AI model designed to tackle everyday questions, including those related to health. The company claims Muse Spark can help by analyzing lab results, fitness tracker numbers, or glucose monitor readings. Users are encouraged to input their personal health data so the AI can identify trends, flag patterns, and create visualizations.
Privacy Risks and Data Security
Although personalized recommendations might sound appealing, experts raise serious concerns. Patient health data shared with Muse Spark is not protected under HIPAA—the US law that safeguards sensitive health information in clinical settings. According to Duke University’s Monica Agrawal, uploading raw medical data to these platforms exposes users to significant privacy risks, since big tech AI tools are not held to the same security and privacy standards as hospitals or physicians.
Meta's privacy policy states that user-provided information may help train future AI models. While Meta says it does not use health data for advertising, the way this data is stored and handled remains loosely regulated compared to healthcare institutions.
Unreliable Health Insights
Healthcare professionals like Dr. Gauri Agarwal recommend using AI for general information or to help prepare questions for your doctor, but not for analyzing raw health data. She points out that users have little transparency or control over how AI platforms store and process their information. Experts also warn that AI chatbots can provide misleading or even dangerous health advice, especially if users ask for extreme solutions or nudge the conversation in risky directions. Muse Spark, for example, was willing to suggest unhealthy meal plans when prompted, overlooking the dangers of conditions like eating disorders.
Should You Trust AI with Your Health Data?
Despite reassurances that Muse Spark is just a helpful assistant and not a doctor, its ability to process and make judgments about medical data is limited. AI-generated reports could lull users into a false sense of security, potentially leading them to share more sensitive information or act on problematic advice.
The bottom line: while digital tools can help you better understand health trends, be cautious about sharing raw medical data with AI chatbots, especially when the privacy standards and accuracy of such tools have not been independently vetted.
For a deeper look at this topic, read the original article by Reece Rogers on WIRED.
