AI Can Now Read Your Doctor's Notes — Here's What Actually Changes
OpenAI connected ChatGPT to Epic, the system holding 325 million patients' medical records, making AI a real clinical tool for the first time.
- 325M
- patients in Epic's system
- 99.1%
- safety rating in physician reviews
- 700,000+
- AI responses reviewed by doctors
- 9
- official health data sources in the plugin
Something big happened in healthcare this week. OpenAI connected ChatGPT to Epic — the computer system that stores medical records for 325 million patients across the US. Starting September 1, doctors at hospitals like UCSF, Cedars-Sinai and HCA Healthcare can open a patient file and ask ChatGPT to summarize it, pull up recent test results or prepare handover notes — without ever switching screens.
The hardest part of AI in healthcare was never the technology. It was getting close to patient data safely. Doctors have always had to open several systems, read pages of notes and piece the story together by hand. Now a HIPAA-compliant AI — one that meets the legal rules for handling medical records — can do that work in seconds. It is read-only, meaning it can look but not write, so the doctor makes every final call.
OpenAI did not just flip a switch. Physicians in 60 countries reviewed more than 700,000 AI responses and rated 99.1% of them safe across 27 different clinical situations. A separate test gave the system a 93% accuracy rating across five different data sources. Those are not small numbers.
For anyone building software for hospitals or clinics, the ground just shifted. ChatGPT is now inside the clinical workflow. The question is no longer 'should we add AI?' — that question has already been answered. The real work is building alongside what is already there. For teams in law, finance or any other regulated field, this is what comes next: when a platform-level AI arrives with a compliance wrapper and safety data, it does not wait.