OpenAI Rolls Out ChatGPT Health to All US Adults, Powered by New GPT-5.6 Sol Model

OpenAI is expanding ChatGPT Health, a feature that lets the chatbot reference a user's personal medical data in any conversation, to all users 18 years and older in the US. The rollout follows a limited test that began in January in a separate portal.

The updated version integrates directly into the main ChatGPT experience rather than living in a standalone tool. Users can connect health data through a new Health tab, linking Apple Health via Apple's HealthKit framework, along with providers such as Kaiser Permanente and One Medical. ChatGPT will also prompt users to confirm current conditions, medications, and family medical history to ensure accurate interpretation of records.

OpenAI claims the underlying model powering these health responses, GPT-5.6 Sol, improves the chatbot's ability to answer health-related questions compared to prior versions. The company has not disclosed specific benchmark scores or technical details distinguishing GPT-5.6 Sol from other models in the GPT-5 line.

Once connected, the Health tab shows a browsable history of records, recent health trends, and suggested prompts. But according to OpenAI, the core function of ChatGPT Health is contextual — the chatbot will automatically reference relevant health data whenever a conversation touches on medical topics, without requiring the user to navigate to the tab first.

OpenAI frames the feature's intended use narrowly: summarizing appointments, interpreting test results, and helping users prepare questions for doctors. The company explicitly states responses should not be treated as diagnosis or treatment. That framing does not prevent users from asking ChatGPT for direct medical advice, and the chatbot can still respond confidently even when incorrect.

On privacy, OpenAI says connected health information is not used to train its foundation models or for ad targeting, regardless of a user's other account settings. Users can require ChatGPT to ask permission each time it wants to reference connected health data, and can disconnect any linked records at any time.

Context: an active lawsuit

The rollout arrives as OpenAI faces a lawsuit from a pastor who alleges ChatGPT gave him medical guidance that led him to forgo treatment for a pulmonary embolism. OpenAI is separately involved in litigation with Apple, a company whose HealthKit framework is now a primary data pipeline for ChatGPT Health. OpenAI has not commented publicly on how either dispute affects the health feature's rollout or design.

What this means

OpenAI is betting that grounding ChatGPT in personal medical records — rather than general health knowledge — makes the tool more useful and, implicitly, safer. But the feature launches into a legal environment that directly tests the opposite claim: whether ChatGPT's medical guidance, however confidently delivered, can cause real harm when treated as authoritative. The privacy safeguards (no ad targeting, no training-data use, per-instance permission) address one set of concerns, but they don't resolve the core tension OpenAI itself acknowledges: a chatbot that summarizes test results one moment can be asked for a diagnosis the next, and nothing in the product stops that shift. For users, the calculus is straightforward — the convenience of a chatbot that knows your medical history has to be weighed against sharing that history with a company currently defending its medical-advice practices in court.