OpenAI connects ChatGPT Health to Epic's EHR for read-only clinical use

OpenAI brings ChatGPT Health into Epic's records

OpenAI has integrated ChatGPT Health with Epic's electronic health record system, giving the assistant a read-only window into data covering more than 325 million patients. According to TechCrunch, clinicians can now import patient data into ChatGPT to ask questions, and in some systems the chatbot will be embedded directly within EHR workflows.

What clinicians can do

With the integration, clinicians can pull in appointment notes, laboratory results, medications and specialist documentation and ask ChatGPT to summarize them. They can also review patient history, spot changes over time and prepare for upcoming visits. In certain deployments, doctors can run a pre-visit review or build a clinical timeline without leaving the patient chart.

OpenAI was explicit that the connection is read-only: the AI does not write anything back to the health record. That keeps the integrity of the medical record out of the model's hands, at least for now, and frames ChatGPT as a reading and summarizing aid rather than a documenting agent.

Public health data plugin and compliance

Alongside the Epic connection, OpenAI launched a Healthcare Public Data plugin that can fetch information from ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed. TechCrunch reports it is intended to help healthcare workers synthesize trial eligibility criteria, medication identifiers, coverage policy versions and provider records.

The company is also allowing organizations with a Business Associate Agreement to use ChatGPT Work, Codex, apps and connectors in their workspace for compliant workflows — the contractual arrangement health systems typically need before a vendor can process patient data.

The safety backdrop

The announcement arrives amid ongoing scrutiny. TechCrunch notes that a Florida-based pastor sued OpenAI days before this rollout, alleging that a ChatGPT recommendation nearly proved fatal, and that in May the family of a user filed suit over alleged dosage advice. OpenAI has continued to maintain that its AI is not suitable for diagnosis or treatment.

The company's own evaluation, as reported by TechCrunch, gathered over 4,300 responses from physicians across 27 clinical use cases — including pre-visit review, clinical timelines, medication review and handoff summaries — and found 99.1% of responses were safe. Even a small fraction of unsafe outputs can lead to harmful results for patients, which is exactly what the remaining percentage represents at scale.

For context, OpenAI brought ChatGPT Health to all U.S. consumers a month before this announcement and has said people ask roughly 300 million health-related questions to the assistant every week.

Why it matters

Placing an AI assistant inside the EHR — the software backbone of everyday clinical work — moves generative AI from consumer curiosity into professional medical infrastructure. With Epic's records covering hundreds of millions of patients, this is one of the largest real-world tests of assistant-style AI in healthcare, and the read-only constraint shows OpenAI anchoring the first step firmly on the cautious side of the line: summarizing and orienting, not authoring. If clinicians find the tool saves meaningful time, pressure will likely grow to expand toward write access and deeper automation; if errors surface in a clinical setting, the lawsuits already on file suggest the legal and regulatory response will be swift. Either way, the boundary between general-purpose AI tools and regulated clinical software is now being redrawn from the inside.