Can nurses use ChatGPT to write patient emails? The privacy risks, and a safer way

October 30, 2026

Nurses, case managers and care coordinators write a lot: discharge follow-ups, transfer requests, messages to families and physicians. AI assistants make that writing faster and clearer, and many clinicians already use them.

The catch is what happens to the text.

What happens when you paste an email into ChatGPT

Online assistants such as ChatGPT, Gemini or Claude run on the provider’s servers. When you paste an email:

In a widely cited JAMA commentary, physicians warned that health systems whose providers use ChatGPT carelessly with patient data could be exposing themselves to HIPAA violations (Becker’s Hospital Review). Outside the US, privacy laws such as Quebec’s Law 25 raise similar questions.

This article isn’t legal advice: follow your organization’s policy, and ask your privacy officer when in doubt.

“I’ll just remove the name”

De-identifying text by hand is harder than it looks. The same commentary lists what has to go: names and nicknames, small geographic details, and admission and discharge dates. A room number, a rare condition or an unusual situation can still point to one person. And when you’re busy, it’s easy to miss a medical record number at the bottom of a message.

A safer way: AI that runs on your computer

The problem disappears if the text never leaves your computer. That’s the idea behind Confi:

You still get what makes AI useful (clearer wording, a more professional tone, shorter messages, English ↔ French translation) without wondering where patient information goes.

Good habits, whatever the tool

  1. Include the minimum: a record number rather than a full name and date of birth.
  2. Review before sending: AI can make mistakes, even on your device.
  3. Use secure channels for sensitive details when your organization provides them.
  4. Follow your organization’s AI policy.

Try Confi for free