OpenAI adds Epic health-record integration to ChatGPT for Healthcare

OpenAI said on September 1 that healthcare organizations can connect authorized Epic electronic health record data to ChatGPT for Healthcare, giving clinicians a route to pull patient chart context into the assistant.

The company said the connection lets a clinician review a patient’s history, identify what has changed and prepare for an appointment inside ChatGPT, and that in supported deployments the assistant can be embedded directly in the EHR layout. Access runs through organization-managed ChatGPT for Healthcare deployments and eligible configured ChatGPT Enterprise deployments, not individual accounts.

OpenAI said ChatGPT can summarize relevant developments from connected records and point clinicians back to the supporting information in the chart. The healthcare-specific tools join a product line that recently added GPT-5.6 across ChatGPT, the API and Codex.

Alongside the record connection, OpenAI introduced a Healthcare Public Data plugin linking ChatGPT to nine official public healthcare sources, naming ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed among them. Eligible U.S. users of ChatGPT for Clinicians can install it. The plugin provides direct, structured access through dedicated connectors, allowing ChatGPT to query records, fields, identifiers and versions in those datasets.

On governance, OpenAI said ChatGPT for Healthcare includes role-based access, single sign-on and audit logs, and that customers with an applicable Business Associate Agreement can use covered workspace tools for HIPAA-compliant workflows.

Two evaluation figures accompanied the launch, both produced by OpenAI. The company said physicians rated 99.1% of responses safe across 4,363 ratings spanning 27 clinical use cases that involved connected EHR context, and that more than 93% of responses for each of five tested connected data sources were rated good or better for accuracy in a two-round physician evaluation. The announcement does not provide the evaluation protocols, the distribution of failures or any external validation, and it reports no patient outcomes, time savings or measured effect on clinician workload.

UCSF Health is named as a pilot partner, and the announcement does not present clinical-outcome results from that pilot. “We’re also engaging frontline teams to validate these capabilities in practice and help shape where they can add the most value,” said Suresh Gunasekaran, president and chief executive of UCSF Health.