OpenAI adds Epic EHR context and public healthcare data sources to ChatGPT

OpenAI announced on September 1 that healthcare organizations can connect authorized electronic health record context from Epic to ChatGPT for Healthcare, alongside a new Healthcare Public Data plugin that gives teams structured access to official medical and healthcare datasets. The move brings ChatGPT closer to the systems clinicians, researchers and administrative teams already use, while putting the product squarely into one of the most regulated and high-stakes areas for enterprise AI.

The Epic integration is designed to help approved users review patient context without manually searching across appointment notes, lab results, medication lists and specialist documentation. OpenAI said ChatGPT can bring together relevant information from the authorized patient record, summarize important changes and point back to supporting chart information. The company described two deployment patterns: bringing EHR context into ChatGPT, and integrating ChatGPT directly inside supported EHR workflows so teams can use AI assistance without leaving the patient chart. OpenAI emphasized that the experience is intended to complement existing workflows rather than replace clinical judgment.

The public-data side is broader. OpenAI said the Healthcare Public Data plugin connects to nine official public healthcare sources, including PubMed, DailyMed, RxNorm, ClinicalTrials.gov and CMS Coverage. That matters because healthcare work often depends on precise identifiers, versions, labels, trial criteria and policy language. A research team might compare recruiting trials, a pharmacy team might verify the current label and warnings for a medication, and a population-health group might combine research, trial information and Medicare coverage details when designing a program. By making those sources queryable in a governed workspace, OpenAI is trying to reduce the friction of switching across databases while still keeping answers anchored to official records.

The company also released evaluation figures aimed at addressing the obvious reliability concern. OpenAI said it works with hundreds of physicians across 60 countries, 49 languages and 26 specialties to define, measure and improve health responses in ChatGPT. Those physicians have reviewed more than 700,000 model responses, according to the company. For connected EHR context, physicians evaluated 27 clinical use cases, including pre-visit review, clinical timelines, medication review and handoff summaries. Across 4,363 ratings, OpenAI said 99.1% of responses were rated safe. In another evaluation using large U.S. healthcare datasets, more than 93% of responses for each of five tested connected data sources were rated as having good or better accuracy.

Those numbers are encouraging, but they do not remove the need for governance. EHR data includes sensitive patient information, and clinical work is filled with edge cases where an incomplete summary or misplaced emphasis can matter. OpenAI’s announcement repeatedly frames ChatGPT for Healthcare as a governed workspace with enterprise controls such as role-based access, single sign-on and audit logs. It also says customers with an applicable Business Associate Agreement can use ChatGPT Work, Codex, apps and plugins in workflows designed to support HIPAA compliance. Individual clinicians in the United States may be eligible to install the Healthcare Public Data plugin, but the EHR integration is not available for individual accounts.

The broader industry signal is clear: AI assistants are moving from answering general medical questions toward operating on authorized clinical and operational context. That could save time in chart review, research synthesis and administrative preparation, but only if hospitals and clinics keep strong review practices around the model’s outputs. OpenAI’s launch shows how the next phase of healthcare AI may be less about standalone chatbots and more about connecting models to trusted records, official datasets and the software environments where care teams already work.