OpenAI links ChatGPT Health to Epic’s EHR covering 325 million patients

OpenAI announced today that ChatGPT Health is now integrated with Epic’s electronic health-record (EHR) system, which holds data for more than 325 million patients. The integration lets physicians import patient information and pose questions to the model with read-only access; the model does not write back to the record. In some deployments the chat appears directly inside the EHR workflow, allowing doctors to review before a visit and build a clinical timeline without leaving the patient file.

The new tool enables clinicians to pull meeting notes, lab results, medication lists and specialist documentation and summarize them easily. It also lets users scan a patient’s history, spot changes and prepare for upcoming appointments. According to the company, this functionality is already available in a subset of organizations that have deployed the connection.

Alongside the Epic integration, OpenAI is releasing a public-medical-information plugin that draws from ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed. The plugin is designed to help medical teams synthesize data for trial-eligibility criteria, drug identifiers, coverage-policy versions and supplier records, all within the same interface.

The firm is also widening access to organizations that have signed a Business Associate Agreement. Those partners will be able to use ChatGPT Work, Codex, applications and connectors in their own environments for regulatory-compliant processes, a step that signals an effort to turn the platform into a broader health-industry infrastructure.

Last month OpenAI launched ChatGPT Health to all consumers in the United States and reported 300 million health queries per week. The company continues to state that the model is not suitable for diagnosis or treatment, a disclaimer that appears in every official document.

OpenAI said it has gathered more than 4,300 physician responses across 27 clinical scenarios—including pre-visit review, clinical timelines, medication review and handoff summaries—and found that 99.1 % of the answers were safe. The company acknowledges that even a small number of unsafe answers could cause real harm.

Two lawsuits are pending. A Florida priest sued days before the launch, alleging that a recommendation nearly cost him his life, and in May family members filed a suit over incorrect dosage advice that led to death.