PALO ALTO, September 3 — OpenAI has switched on a direct integration between ChatGPT Health and Epic, the hospital software platform that holds records for more than 325 million patients, the company announced on September 1. The deployment is live, embedding ChatGPT inside Epic so clinicians can pull up AI-assisted chart review without leaving the system they already work in. Under the hood, the technical reality is deliberately constrained.
Physicians can import and query patient records within Epic workflows, but access is read-only — ChatGPT cannot write back to the chart. That guardrail matters because the liability terrain is already contested.
Reuters has reported a lawsuit from the family of a 19-year-old who died after allegedly receiving dangerous drug advice from ChatGPT, and the BBC has covered a separate suit from a Florida pastor who claims the AI discouraged him from seeking medical care after repeated symptom discussions. Organisations with a Business Associate Agreement in place can also deploy ChatGPT Work and Codex; that contract is the HIPAA-relevant threshold for compliant workflows.
The safety numbers, and the volume problem
OpenAI’s internal figures are encouraging, but raw query volume complicates the picture. The company reviewed more than 4,300 physician responses across 27 clinical use cases and found 99.1% were rated safe, per TechCrunch’s report. A separate review of 6,924 conversations rated 99.6% safe and accurate, according to OpenAI.
Yet OpenAI reportedly recorded more than 300 million health-related queries per week as of July 2026. Apply the 0.9% error rate to that volume and the arithmetic yields millions of potential clinical missteps — a patient-safety variable worth tracking with the same rigour as any other clinical metric.
Disclaimers and oversight
OpenAI continues to state that ChatGPT is “not intended for diagnosis or treatment.” That disclaimer does real work, given the pending litigation. The read-only design is a key safety feature, but the integration’s scope and speed of deployment raise questions about oversight — and about the need for continuous monitoring of AI performance in clinical settings as the tool spreads across Epic’s installed base.
