Quick answer
OpenAI has released two distinct healthcare plugins for eligible ChatGPT and Codex workspaces. The Epic plugin gives approved organizational users read-only access to patient information they are already authorized to view in Epic. Healthcare Public Data searches nine official public sources for research, trials, medication information, Medicare data, facility measures, and provider identifiers; it does not access patient charts. Both plugins require appropriate workspace and app controls, and the public-data tools should not receive protected health information. OpenAI's first-party evaluations show strong physician ratings, but clinicians remain responsible for checking source records and making care decisions.
OpenAI launched two separate healthcare plugins for ChatGPT and Codex
The Epic plugin connects an approved organizational workspace to authorized patient-chart information. Healthcare Public Data searches official public datasets and does not retrieve patient records. OpenAI has made both available to eligible ChatGPT for Healthcare and HIPAA-enabled ChatGPT Enterprise workspaces in ChatGPT and Codex. Keeping the integrations separate matters because they use different data, permissions, setup, and privacy boundaries.
Sources: OpenAI, OpenAI, OpenAI Help Center, OpenAI Help Center
Epic access is read-only and follows existing chart permissions
A workspace administrator and Epic administrator must configure the required EHR app, Epic FHIR endpoint, OAuth application, and approved scopes. Each clinician then connects an individual Epic account. The plugin can retrieve permitted notes, medications, conditions, encounters, and laboratory results, but it cannot update the medical record, place orders, message patients, or bypass existing patient-chart permissions. Users must still review the underlying chart and relevant dates before relying on a summary.
Sources: OpenAI Help Center, OpenAI
Healthcare Public Data searches nine official sources without opening patient charts
The public-data plugin groups nine read-only apps: PubMed, ClinicalTrials.gov, DailyMed, RxNorm, openFDA, CMS Coverage, CMS Open Data, Medicare Care Compare, and the NPI Registry. These sources cover research, trials, medication labels and identifiers, public FDA safety information, Medicare coverage and payment data, facility measures, and provider identifiers. Each source has its own limits; for example, an adverse-event report does not prove causation, an NPI does not establish current licensure, and a public coverage policy does not determine an individual's benefits.
Sources: OpenAI Help Center, OpenAI
Do not send protected health information to the public-data apps
OpenAI explicitly tells users not to include patient names, dates of birth, medical-record numbers, Medicare or member numbers, contact details, or other protected health information in searches sent to public sources. A workspace Business Associate Agreement does not automatically authorize an external public-data provider to receive that information. Organizations should review their agreements, connected services, retention requirements, data-residency rules, app permissions, and compliance logs before approving a workflow.
Sources: OpenAI Help Center, OpenAI
Availability differs for organizations and individual clinicians
Eligible ChatGPT for Healthcare and HIPAA-enabled Enterprise workspaces can use both plugins after the required administrator setup. Eligible U.S. users of ChatGPT for Clinicians can install Healthcare Public Data, but the Epic integration is not available for individual clinician accounts. Plugin installation, workspace permission, individual app access, Epic configuration, and user sign-in are separate steps; seeing a plugin in the directory does not mean every included app or patient record is available.
Sources: OpenAI Help Center, OpenAI Help Center, OpenAI
OpenAI's physician evaluations are encouraging but not a clinical guarantee
OpenAI says physicians evaluated connected EHR responses across 27 clinical use cases and rated 99.1% of 4,363 responses safe. In a separate evaluation of nuanced questions using five connected U.S. healthcare data sources, more than 93% of responses for each source received a rating of good or better for accuracy. These are first-party evaluations with defined tasks and rating methods. They do not establish error-free performance, diagnostic validity, suitability for every specialty, or a replacement for clinical review.
Sources: OpenAI
This is a regulated ChatGPT product update, not a Chat AI feature announcement
The release concerns plugins for eligible OpenAI workspaces and individual ChatGPT for Clinicians accounts. It does not announce an Epic connection, Healthcare Public Data plugin, or protected-health-information workflow in Chat AI. Healthcare data access is product- and configuration-specific, so teams should verify the exact workspace, agreements, permissions, and connected services rather than assuming that one assistant's integration carries over to another app.
Sources: OpenAI, OpenAI Help Center
Frequently asked questions
What readers usually ask
Can ChatGPT now read Epic patient records?
Approved organizational users can connect their own Epic accounts in eligible configured workspaces and review patient information they already have permission to access. The connection is read-only and does not expand existing chart permissions.
Can the Epic plugin write to the medical record?
No. OpenAI says the current integration cannot update records, place orders, message patients, or override patient-chart permissions.
Which sources are included in Healthcare Public Data?
PubMed, ClinicalTrials.gov, DailyMed, RxNorm, openFDA, CMS Coverage, CMS Open Data, Medicare Care Compare, and the NPI Registry.
Does Healthcare Public Data access patient charts?
No. It searches public healthcare information. Epic is a separate plugin for authorized patient-chart access in approved organizational workspaces.
Can individual clinicians use both plugins?
Eligible U.S. ChatGPT for Clinicians users can install Healthcare Public Data. The Epic plugin is not available for individual clinician accounts.
Should protected health information be entered into the public-data plugin?
No. OpenAI instructs users not to include protected health information or patient identifiers in searches sent to public healthcare sources.
Are these healthcare plugins available in Chat AI?
No Chat AI availability is announced or verified. This release applies to specific OpenAI workspaces and eligible ChatGPT for Clinicians accounts.
Evidence