ChatGPT risks health privacy by storing sensitive personal data
Vulnerabilities in Personalization: Assessing Health Privacy Risks in ChatGPT Logs and Memory
Human-Computer InteractionComputers and Society
Summary
Many people share personal health details when chatting with ChatGPT, but this study found that the AI stores much of this sensitive information without users knowing. The authors analyzed over 179,000 chats from four countries and found that some conversations included risky details like exact locations and stigmatized conditions. ChatGPT also turns temporary symptoms into permanent health profiles, which can increase privacy risks. The authors suggest ways to give users more control over what the AI remembers.
What this means in practice
- •For ai product managers: Improve features that manage user health data to provide clearer consent and limit storing sensitive information without explicit prompts.
- •For health app developers: Design chatbot memory systems that preserve user privacy by avoiding synthesis of sensitive temporary health details into permanent profiles.
Authors
S M Mehedi Zaman, Md Mozammel Hoque
Abstract
As conversational LLMs become deeply embedded in daily life, users frequently disclose sensitive personal health information during routine interactions. We present a large-scale computational audit analyzing 179,057 conversations across India, Nigeria, Brazil, and Pakistan (N = 1,057) to evaluate personal health disclosures and background memory synthesis in ChatGPT. We find that 21.31% of audited conversations contain personal health data, with 3.62% posing high-to-extreme privacy risks involving stigmatized conditions, direct identifiers, and precise locations. When evaluating the memory entries of ChatGPT, we uncover a stark disconnect between corporate framing and system behavior: over 95% of profile entries are implicitly extracted without explicit user prompts or consent. Furthermore, background memory synthesis selectively condenses temporary, symptom-level disclosures into permanent diagnostic traits, stripping contextual integrity and amplifying re-identification risks. We conclude with sociotechnical design guidelines to restore user agency and consent-driven boundaries in stateful AI systems.