OpenAI Bridges ChatGPT to Epic EHR, Letting Doctors Access Patient Data Through AI

OpenAI has announced a direct integration between ChatGPT Health and Epic's electronic health record platform, which manages information for more than 325 million patients. The move allows medical professionals to import patient records directly into ChatGPT and leverage AI to analyze and retrieve health information efficiently.

What Clinicians Can Now Do#

Through the integration, doctors gain AI-assisted access to a patient's appointment summaries, test results, medication lists, and reports from specialists. The tool can rapidly synthesize this information, helping clinicians spot trends in a patient's medical history, note significant changes, and prepare for upcoming visits. In selected hospital and clinic deployments, clinicians can invoke ChatGPT without leaving the patient record—enabling pre-appointment reviews and structured clinical timelines all within existing EHR workflows.

OpenAI emphasizes that the system operates on a read-only basis, meaning the AI cannot alter or send information back into patient records.

New Healthcare Data Plugin#

Alongside the Epic integration, OpenAI introduced a Healthcare Public Data plugin that taps into several authoritative medical databases, including ClinicalTrials.gov, CMS Coverage information, RxNorm, DailyMed, and PubMed. This tool helps clinicians identify trial candidates, cross-reference medication names, check insurance coverage rules, and locate provider documentation—all within the AI's conversational interface.

Organizations operating under a Business Associate Agreement can also now deploy ChatGPT Work, Codex, apps, and connectors in HIPAA-compliant workflows.

The Broader Health AI Context#

OpenAI rolled out ChatGPT Health to U.S. consumers last month, noting that the tool fields around 300 million health-related questions weekly. The company has stressed repeatedly that AI should not be used for diagnosis or treatment—only as an informational aid.

Yet safety concerns linger. OpenAI studied over 4,300 physician responses across 27 clinical scenarios and reported a 99.1% safety rate. However, even rare unsafe outputs can pose serious risks. Recently, a Florida pastor sued OpenAI, claiming ChatGPT gave him life-threatening guidance, and in May, family members of another user pursued litigation, alleging the system provided dangerous dosage advice.