Safety, privacy, and what to never type
The four risks
- Hallucination: confident but wrong answers.
- Privacy and PHI exposure: assume pasted content may be stored or reviewed unless you are on an appropriate business or enterprise plan with the right settings.
- Over-reliance: if you stop thinking, the AI's mistakes become yours.
- Outdated information: drug guidance, regulations, and protocols change.
Never type
- Full patient names
- Phone numbers, addresses, Aadhaar numbers, or insurance IDs
- Photos of ID cards, prescriptions with names, or lab reports with identifiers
- Staff salaries or disciplinary details with names
- Anything you would not be comfortable seeing on a billboard
Safer to type
- Anonymized clinical scenarios with ages rounded
- General questions about conditions, drugs, or protocols that you will verify
- Drafts of policies, letters, and SOPs
- Your own writing that you want improved
Personal AI hygiene checklist
Before pasting anything into a chat, ask:
- Did I remove all names and identifiers?
- Am I treating the output as a draft, not a final answer?
- Will I verify clinical facts before acting?
- Have I read the output critically?
- If a family member saw this chat, would I be comfortable?
Learning Resources
Study Assistant
External links
Hi! I'm your learning assistant. Use the external links below to explore "Safety, privacy, and what to never type" safely in ChatGPT or Claude.
Need Help?
Join our expert-led forum to discuss case studies with fellow clinicians.
Study Assistant
External links
Hi! I'm your learning assistant. Use the external links below to explore "Safety, privacy, and what to never type" safely in ChatGPT or Claude.
Need Help?
Join our expert-led forum to discuss case studies with fellow clinicians.
