Meta’s New AI Asked for My Raw Health Data—and Gave Me Terrible Advice

Miami: A Wired reporter tested Meta’s new AI health assistant called Muse Spark and discovered it gave terrible advice after asking for personal health information. The AI chatbot wanted the reporter to upload raw medical data like lab reports.

Medical experts are worried about this new technology. Doctors say people should not trust AI with their private health information. “I certainly wouldn’t connect my own health information to a service I can’t fully control,” says Dr. Gauri Agarwal from the University of Miami.

The AI claimed it was like a medical school professor, not a real doctor. But when the reporter asked about losing weight, the chatbot suggested extreme fasting. Even when told to eat only 500 calories a day, five days a week, Meta AI created a meal plan that could make someone very sick.

The AI can be easily tricked into giving dangerous health advice. Experts warn that people asking questions might unintentionally lead the AI to wrong answers. “A model might take information as truth without questioning it,” explains Dr. Agarwal.

Meta says people control what information they share with the AI. But privacy experts remain concerned about where personal health data goes and who can see it. The bot sometimes told users to remove personal details before uploading lab results, but not always.

Medical professionals say these AI tools might seem helpful when doctor visits are expensive. But they warn that replacing real doctor visits with AI chatbots could be dangerous. “Running into that without due diligence is dangerous,” says Kenneth Goodman from the University of Miami’s Bioethics Institute.

The reporters found that even simple health questions could lead to harmful advice from Meta’s AI system. Experts strongly recommend people stick to asking their real doctors instead of relying on artificial intelligence for health guidance.