OpenAI has added new healthcare-data connections for eligible organizational customers, including a read-only Epic integration that can bring authorized patient context into ChatGPT for Healthcare and Codex. The September 1 release is aimed at clinicians and healthcare teams working inside approved enterprise environments, not at giving ordinary consumer ChatGPT accounts access to patient charts.
The Epic plugin can surface information a clinician already has permission to review, such as clinical notes, medications, conditions, encounters and laboratory results. OpenAI says access follows the user's existing Epic and patient-chart permissions. An organization must configure the required Epic app, and each clinician signs in with an individual Epic account.
Read-only is the key boundary
The integration is currently read-only. OpenAI's product documentation says it cannot update medical records, place orders, message patients or override existing chart permissions. That distinction matters because the feature is designed to help clinicians review and synthesize authorized information rather than act as an autonomous charting or care-delivery system.
OpenAI gives examples such as reviewing what has changed since a previous visit, checking recent laboratory results, identifying medication changes and preparing for an appointment. The company says ChatGPT points back to supporting chart information so clinicians can inspect the underlying record.
Clinicians remain responsible for reviewing source records and making care decisions. The feature should therefore be understood as a governed information-access layer inside eligible healthcare workspaces, not as a replacement for clinical judgment.
A separate plugin covers public healthcare data
OpenAI also introduced Healthcare Public Data, a plugin that searches nine official public healthcare sources. The company's examples include ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed. That plugin is also read-only, but it is separate from Epic and does not retrieve patient charts.
The separation is important for privacy and workflow. Public-data queries can be used for research, medication information, coverage material and clinical-trial records, while Epic is the route for authorized organizational patient context. OpenAI warns users not to include protected health information in searches sent to public healthcare sources.
Availability depends on the product and workspace. OpenAI says the Epic plugin is for approved ChatGPT for Healthcare and HIPAA-enabled ChatGPT Enterprise workspaces, subject to an organization's rollout and configuration. Healthcare Public Data also has its own eligibility and workspace controls.
For healthcare teams, the practical change is that more of the information they already use can be reviewed from a governed ChatGPT workspace without treating every data source as the same thing. Epic remains permission-bound patient context. Public Data remains a collection of official external sources. Neither connection changes the underlying responsibility to verify information and make human clinical decisions.
That makes this a meaningful integration release, but a deliberately constrained one. The useful headline is not that ChatGPT can now edit a patient chart. It cannot. The product is being positioned to help authorized teams find and understand information while keeping the source systems and existing access controls in charge.
For administrators, that governance model may be as important as the interface. Enabling a plugin, connecting an app and having permission to view a particular patient record are separate controls. The new connection does not collapse those layers into one broad grant of access.
Reporting notes