2026-08-25 · 4 min read

AI in healthcare: the tools without the confidentiality breach

Where AI already works in healthcare — and where the risk is

The everyday uses aren't futuristic diagnostics; they're paperwork: visit summaries, discharge letters, insurer correspondence, translation and chart summarization. That is exactly where the risk sits — the input is a document carrying a name, ID, diagnoses and treatments, and the destination is an external server.

The numbers explain the sensitivity: per IBM Cost of a Data Breach 2024, a healthcare breach averages $9.77 million — the most expensive of all industries, for the 14th consecutive year. Medical data can't be "replaced" like a credit card.

What must never enter a public AI tool in a medical setting

  1. Identifying patient details — name, ID number, date of birth, address, phone.
  2. Diagnoses and treatments next to any identifier — "a 47-year-old patient from town X with diagnosis Y" identifies too.
  3. Scanned documents — referrals, summaries and photographed records carry identifiers on every line.
  4. Staff information — employee details, clinical-system passwords, permission structures.
  5. Non-de-identified research data — study files with participant identifiers.

How to still use AI on medical documents

  1. Set policy: which tools are approved, business-tier only, no training on your data.
  2. Require anonymization before every paste: the patient's name becomes PERSON_001, the ID becomes ID_001 — diagnoses and clinical course stay intact, so the AI's summary stays accurate.
  3. Prefer local processing: a tool that works in the browser or on the device — the original document never leaves the institution.
  4. Clean Word metadata — the Author field exposes the drafting clinician's name.
  5. Train the staff: most leaks are a visit summary pasted into a chat in good faith.

The principle matches the general list of what never to paste into ChatGPT — only the bar here is higher: with medical data, even a single leak is a reportable incident and real harm to a person.

Frequently asked questions

May a physician use ChatGPT to draft letters?

Yes — as long as the input contains no identifying patient details. An anonymized document exposes no identifiable person's medical information, so entering it isn't a confidentiality breach.

Is removing just the name enough to de-identify a medical document?

No. An ID number, date of birth, a small town, a rare diagnosis, or a combination of them identifies a person without any name. Real anonymization also handles indirect identifiers and file metadata.

Does an internal hospital AI solve the problem?

It reduces the outbound risk, but not the internal permission and logging questions. Even against an internal system, minimizing identifying data in the input is good practice — and shrinks any future leak.

What does HIPAA say about AI tools?

PHI may only be shared with vendors under a Business Associate Agreement. A public consumer AI tool has no BAA — so PHI must not enter it. De-identified data, per the HIPAA standard, is outside that restriction.