OpenAI is moving ChatGPT Health deeper into clinical work by connecting it with Epic’s electronic health record system. The integration is designed to let clinicians bring patient information into ChatGPT Health and use AI to review, summarize, and question that information during care preparation.
The update matters because Epic’s EHR system holds data for over 325 million patients. OpenAI also said that in certain systems, ChatGPT will be integrated directly within EHR workflows, giving clinicians access without leaving the patient chart.
What the Epic integration enables
With the new ChatGPT Health and Epic connection, clinicians can access several types of patient information. The source article lists appointment notes, laboratory results, medications, and specialist documentation as examples of data that can be imported for review.
The practical goal is not to replace a clinician’s judgment. It is to make large amounts of patient context easier to scan and organize. Clinicians can use ChatGPT to summarize records, review patient history, identify changes, and prepare for future appointments.
OpenAI also said some deployments will allow clinicians to access ChatGPT directly from a patient chart. In those settings, the tool can support pre-visit review and clinical timelines without requiring the clinician to switch out of the EHR workflow.
Read-only access is a key boundary
OpenAI specified that the integration provides read-only access to health records. That means AI can read imported patient data for analysis and summarization, but it does not write anything back into the record.
This distinction is important in clinical workflows. A summary or timeline can help a healthcare worker review information more quickly, but the medical record itself remains unchanged by the AI system. Any documentation or action still sits outside the described write-back capabilities of the integration.
The source also makes clear that OpenAI has maintained a limit around medical use: AI is not suitable for diagnosis or treatment. That caution remains relevant even as ChatGPT Health gains more direct access to medical context through Epic.
A new public health data plugin
Alongside the Epic integration, OpenAI is adding a Healthcare Public Data plugin. The plugin can retrieve information from ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed.
OpenAI said the plugin can help healthcare workers synthesize public data across several healthcare tasks. Those include trial eligibility criteria, medication identifiers, coverage policy versions, and provider records.
Taken together, the Epic integration and the Healthcare Public Data plugin point to two different kinds of information use. One draws from patient-specific records in the EHR. The other draws from public healthcare data sources that can support research, coverage, medication, and provider-related questions.
Compliance and workplace access
OpenAI said organizations with a Business Associate Agreement can use ChatGPT Work, Codex, apps, and connectors in their workspace for compliant workflows. That expands the environment in which healthcare organizations can apply OpenAI tools while working under that agreement.
The announcement follows another recent health-related rollout. Last month, OpenAI rolled out ChatGPT for health to all U.S. consumers. At that time, the company said people were asking 300 million health-related queries to ChatGPT every week.
That consumer usage shows why health has become a major area for ChatGPT. But the clinical integration described here is different from consumer health queries. It is aimed at clinicians working with patient charts, pre-visit review, clinical timelines, medication review, and handoff summaries.
Safety remains the central question
OpenAI said it collected over 4,300 responses from physicians across 27 clinical use cases. Those use cases included pre-visit review, clinical timelines, medication review, and handoff summaries. The company said it found that 99.1% of responses were safe.
Even with that figure, the source article notes the risk clearly: a small number of unsafe answers can still create harmful outcomes for people. That concern becomes more serious when AI is used near patient information, medication context, or clinical decision-making workflows.
The article also notes recent legal pressure around health-related AI responses. A Florida-based pastor sued OpenAI, alleging that ChatGPT gave him a near-fatal recommendation. The company was also sued in May by family members of a user who blamed ChatGPT for giving wrongful advice related to dosage.
For clinicians, the central takeaway is that ChatGPT Health is becoming more capable at organizing and surfacing information, especially through Epic and public healthcare data sources. But the same source makes clear that AI should not be treated as a tool for diagnosis or treatment, and the read-only design reflects that boundary.