OpenAI disclosed that ChatGPT Health has been integrated with Epic’s electronic health records system, allowing doctors to import patient data and use AI to query and organize medical records in compliance with regulations. With Epic serving over 325 million patients, this integration signifies ChatGPT’s further entry into mainstream U.S. healthcare workflows.
Can read medical records but cannot write back
According to OpenAI, doctors can use ChatGPT to access appointment records, lab results, medication information, and specialist documentation, and quickly generate summaries. The system can also help organize patients’ medical histories, identify changes, and prepare for upcoming visits.
In some deployments, doctors can directly invoke ChatGPT without leaving the patient’s chart page to review pre-visit information and organize clinical timelines. OpenAI emphasizes that this integration currently provides read-only access; the AI does not write content back into the electronic health record system.
Add new healthcare public data plugin
In addition to medical record integration, OpenAI has launched the Healthcare Public Data plugin, which retrieves information from multiple public medical databases, including ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed.
The company states that this plugin can integrate information such as trial eligibility criteria, drug identifiers, insurance coverage policies, and healthcare provider records, helping medical professionals quickly retrieve and summarize data.
Expand the scope of the compliance workspace
OpenAI also stated that it has permitted organizations that have signed a Business Associate Agreement to use ChatGPT Work, Codex, apps, and connectors within their workspaces to support compliant healthcare processes.
The company also disclosed that it collected over 4,300 physician feedback responses across 27 clinical use cases, including pre-visit review, clinical timelines, medication review, and handoff summaries. Of these responses, 99.1% were evaluated as safe.
However, OpenAI also noted that even a low percentage of unsafe responses could still cause real harm in medical scenarios, meaning that safety controls will remain a priority in deploying medical AI as it expands access.