OpenAI has just expanded ChatGPT Health’s capabilities by integrating it with Epic’s electronic health record (EHR) system, which stores data for more than 325 million patients. This update allows clinicians to import patient records into ChatGPT Health — including notes, lab results, medications, specialist reports — and then use AI tools to summarize, analyze history, and prepare for upcoming appointments. In select setups, ChatGPT is built right into Epic workflows so care providers can pull in info without leaving a patient’s chart.
The integration is read-only: the AI accesses existing records but does not write, change, or overwrite any patient data. OpenAI is also introducing a Healthcare Public Data plugin. This add-on provides access to external medical resources like ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed. With it, practitioners can cross-reference trial eligibility, medication identifiers, insurance coverage terms, and provider directory records.
This rollout also includes compliance enhancements: larger organizations that have Business Associate Agreements (BAAs) can now deploy ChatGPT Work, Codex, plus associated apps and connectors in workplace settings designed to meet healthcare privacy and regulation needs.
OpenAI officially released ChatGPT for Health to all U.S. users last month. The company reported that the platform is fielding around 300 million health-related questions per week. Still, OpenAI emphasizes that despite all these data integrations, the AI system is not meant for issuing diagnoses or treatment recommendations.
Internal testing covered 27 different use-cases — like medication review, drafting visit summaries, tracking clinical timelines, and performing pre-visit prep — with feedback collected from over 4,300 physicians. According to the company, 99.1% of responses during those trials were safe. But the tiniest number of errors could potentially have serious consequences.
These enhancements arrive amid recent legal challenges. OpenAI is being sued by a Florida pastor who claims ChatGPT offered him nearly lethal advice, and by family members of another user who allege the system gave incorrect dosage information. These claims highlight the risk that comes with deploying AI in health settings at scale.
Why OpenAI’s move matters
Integrating ChatGPT Health with Epic bridges one of the biggest gaps between AI tools and clinical workflow. Medical staff often juggle multiple systems and data sources; giving AI read-only access to full patient records and external data sources reduces friction and could speed up tasks like medication reconciliation, trial enrollment, or care planning.
Still, this setup raises essential questions about safety, accuracy, and accountability. Even with high safety ratings during testing, AI in medicine must handle rare but high-stakes events carefully. Monitoring, transparent audit trails, and strong oversight will be crucial, especially given ongoing legal claims linked to medical advice from AI.
What to watch next
Going forward, it’s worth keeping an eye on how widely this Epic integration is adopted by hospitals and health systems. Also important: how different deployments manage consent and patient data privacy, especially across state lines or institutions with varied security infrastructures. Finally, whether liability frameworks evolve — both legally and ethically — to keep pace with AI tools that increasingly reach inside core facets of clinical decision support.
OpenAI’s latest Epic integration amplifies the potential of AI in healthcare operations. By letting clinicians feed in comprehensive patient data and external resources, the company aims to cut inefficiencies in care coordination and prep. But unless safety, oversight, and legal boundaries are rock-solid, even small missteps could undercut trust in tools that promise so much.
