Some people are feeding years of medical records into chatbots like ChatGPT, despite privacy risks and receiving generalized or inaccurate diagnoses in response
Maggie Astor / New York Times : Bluesky: @anniebkay , @kashhill , and @hypervisible.blacksky.app Bluesky: Annie B Kay / @anniebkay : It's absurd. AI is inaccurate - at least half the time. [embedded post] @kashhill : Dr. ChatGPT is learning even more about people than Dr. Google did www.nytimes.com/2025/12/03/w... @hypervisible.blacksky.app : Legally, “'you're basically waiving any rights that you have with respect to medical privacy,' leaving only the protections that a given company chooses to offer.”
Context & Ripple Effects
This extends a longer pattern of people treating general-purpose chatbots as health advisers, from using AI as a substitute or supplement for mental-health support to clinician use of ChatGPT for patient communication. The new behavior is more consequential because it combines intimate longitudinal records with consumer AI services rather than isolated questions.
The reliability concern is also established in related coverage: medical prompts have produced harmful or inaccurate content, and a later test of health-focused chatbot responses found questionable and inconsistent answers. This report adds the privacy consequence of supplying the underlying records that make advice appear more personalized.
First-order effects
- People who upload complete medical histories trade potentially sensitive health information for responses that may still be generalized or inaccurate; the practical privacy baseline becomes the chatbot provider's policy rather than ordinary medical-care expectations.
- ChatGPT and comparable services are being used in a high-stakes advisory role without the report establishing that their answers are clinically dependable.
Second-order effects
- Healthcare professionals may have to spend more time correcting chatbot-generated interpretations and clarifying which AI tools are appropriate for discussing records.
- The gap between users' expectation of personalized medical guidance and the documented limits of chatbot answers increases pressure for clearer data-use disclosures and health-specific safeguards.
Third-order effects
- If consumers continue moving medical records into general-purpose AI, health-data governance will increasingly have to address consumer platforms, not only conventional care settings.
- The pattern points toward a durable distinction between AI that assists communication and AI that is treated as a diagnostic authority; whether providers can close that gap will shape trust and oversight.
The trend: Consumer chatbots are evolving from low-stakes information tools into de facto health advisers, bringing privacy governance and reliability scrutiny with them.