SAN FRANCISCO, March 8, 2024 — OpenAI has launched ChatGPT Health, a dedicated space within its popular chatbot designed for health and wellness inquiries, and capable of integrating with personal medical records and health apps. The move signals a significant expansion of ChatGPT’s capabilities, tapping into a massive user base already turning to the AI for health-related information.
A Growing Demand for Digital Health Tools
More than 230 million people globally are using ChatGPT each week to ask questions about their health, making it one of the chatbot’s most popular applications.
- ChatGPT Health offers a separate, secure environment for health conversations.
- Users can connect their electronic medical records and wellness apps for personalized responses.
- The system is built with input from over 260 physicians across 60 countries.
- OpenAI emphasizes that ChatGPT Health is a support tool, not a replacement for medical professionals.
ChatGPT Health isn’t just another chatbot feature; it’s a distinct environment within ChatGPT, complete with its own storage, memory, and privacy controls. OpenAI assures users that sensitive health data is kept separate from other conversations and is protected with enhanced security measures. This compartmentalization addresses a key concern surrounding the use of AI in healthcare: data privacy.
The new service allows users to link electronic health records and a growing number of wellness applications. Currently, integrations include Apple Health, b.well’s health data network in the US, and popular consumer apps like Function, MyFitnessPal, Weight Watchers, AllTrails, Instacart, and Peloton. By connecting these sources, ChatGPT Health can tailor its responses to an individual’s specific health profile.
Maintaining Privacy and Security
While health-related conversations remain within the main ChatGPT history for easy access, OpenAI emphasizes that the underlying information stays contained within the Health space. “Health memories” are also kept separate and can be viewed or deleted through personalization settings. All conversations and files are encrypted both while stored and in transit, with ChatGPT Health adding purpose-built encryption specifically for health data. OpenAI confirms that conversations within the Health environment are not used to train its core AI models.
Users have additional control over their data, including the ability to enable multi-factor authentication and disconnect health apps or revoke access to medical records through the settings menu. Any app integrated within ChatGPT Health must meet stringent privacy and security requirements and undergo a dedicated security review.
Initially, ChatGPT Health is rolling out to a limited group of early users. Individuals with ChatGPT Free, Go, Plus, and Pro plans—excluding those in the European Economic Area, Switzerland, and the United Kingdom—can join a waitlist. OpenAI plans to expand access to the web and iOS platforms in the coming weeks. Currently, medical record integrations and some app connections are only available in the US, and Apple Health syncing requires an iOS device.
Clinical Guidance and Development
OpenAI has positioned ChatGPT Health as a tool to support everyday health questions and preparation for medical appointments, explicitly stating it is not intended for diagnosis or treatment. Users can leverage the system to understand recent test results, prepare questions for their doctors, explore diet and exercise plans, or compare healthcare insurance options. The system can also identify trends in connected data, such as changes in cholesterol levels or sleep patterns.
ChatGPT may proactively suggest moving a conversation to the Health space when it detects a health-related topic. While it can draw context from non-health chats—like a recent relocation—information flows one way only. Non-health conversations cannot access data within the Health environment.
Users can interact with ChatGPT Health using photos, files, voice commands, and dictation. They can also utilize integrated apps to create shoppable grocery lists from meal plans or pair activity data with exercise recommendations. Custom instructions within the Health environment allow users to tailor the system’s responses to their preferences regarding tone, focus, and specific topics.
The development of ChatGPT Health involved a two-year collaboration with a network of over 260 practicing physicians from 60 countries. These clinicians provided feedback on model outputs more than 600,000 times, across 30 areas of focus. Their input guided decisions on when to encourage follow-up with a doctor, how to phrase information clearly without oversimplification, and how to handle sensitive situations requiring heightened safety protocols.
OpenAI employs an evaluation framework called HealthBench, based on physician-written guidelines and clinical standards, to assess the system’s performance. This framework prioritizes safety, clarity, appropriate escalation, and respect for individual context over simple accuracy checks.
“This physician-led approach is built directly into the model that powers Health, which is evaluated against clinical standards using HealthBench, an assessment framework we created with input from our network of practicing physicians,” said OpenAI.
OpenAI intends to expand the range of data sources integrated with ChatGPT Health and enhance the insights it can provide as it gathers feedback from early adopters.
