AI in nursing: what works without patient data
Written with AI and checked against the sources by a second AI. As of November 15, 2026. More in the notes on AI.
AI can take some of the writing off your hands in nursing, but not all of it. As soon as real patients, residents or colleagues can be identified, strict rules apply. This guide shows what works well without patient data, what doesn’t belong in a public AI chat (a freely available service with a personal account and no contract with your employer) and whom to ask when in doubt.

Key points
- Under the GDPR, health data is specially protected. Nurses are usually also bound by professional secrecy under Section 203 of the German Criminal Code.
- A lot works without patient data, such as handover templates, training material, letters to relatives without names and plain-language versions of technical texts.
- Diagnoses, medication and records with real data don’t belong in a public AI chat.
- Your facility decides whether to approve an AI and which one. Ask the director of nursing or the data protection officer.
- You are professionally responsible for every result. Check it against the expert standard, your facility’s standard or the doctor’s orders.
In short: a lot works without patient data
Most of the writing in nursing is about people: care notes, handovers, conversations with relatives. That is exactly where caution is needed. But many tasks can be set up so that nobody appears in them. Then an AI like ChatGPT, Claude or Copilot can help without any data about patients or residents leaving the building.
| Works without patient data | Doesn’t belong in it |
|---|---|
| A handover template with empty fields | Today’s handover with real details |
| Training on fall prevention for new colleagues | A particular resident’s fall last night |
| A letter to all relatives about new visiting hours | A letter to the daughter of a particular resident |
| A technical term explained in plain language | A doctor’s letter to summarize |
| Ideas for roster rules | The roster with names and sick notes |
The rule of thumb: if you could send the text to an outside company without anyone being identifiable in it, it can go into an AI chat. Otherwise it stays in the building.
In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson
Why health data is so strictly protected
The EU General Data Protection Regulation (GDPR) counts health data among the special categories of personal data. Under Art. 9, it may only be processed if one of the exceptions listed there applies, for example for care provided by professionals who are bound by confidentiality. Health data is more than the diagnosis: the fact that someone needs help washing or is restless at night also says something about their health.
Then there is professional secrecy. Under Section 203 of the German Criminal Code (Strafgesetzbuch), anyone who discloses another person’s secret without authorization commits an offense if it was entrusted to them, or became known to them, as a member of a health profession with state-regulated training. This usually applies to registered nurses. People who work in care without this training are usually bound to confidentiality by law or by their employment contract.
Finally, your facility has its own rules. Depending on who runs it, state laws or the churches’ own data protection laws may also apply. What your employer allows or forbids for AI applies to you even if a service is technically freely available.
Typing patient data into a public AI chat is a bit like sending the care record to an outside typing service that your facility has no contract with and that nobody has bound to confidentiality.
What works well without patient data
Handover templates
Nadia is a registered nurse on a medical ward. Handovers go differently with every colleague, and important points sometimes get lost. She has the AI draft an empty template the team can try out. Real details are entered later by each nurse in the hospital’s documentation system, not in the chat.
Draft a template for the verbal shift handover on a medical ward.
Divide it into fixed sections, for example condition, changes since the last shift, open tasks, appointments, special points.
For each section, a short guiding question and an empty field.
The template should fit on one A4 page and take two minutes per person to go through.
Don’t include any examples with diagnoses or medication.
At the end, list three questions we should settle as a team before we use the template.Whether the sections fit your facility’s standard is for the team to decide. The AI delivers a draft, not a rule.
Training material
Leo is a geriatric nurse and trains new colleagues. For a short session on fall prevention, he gives the AI his own bullet points and has it turn them into a handout and five quiz questions. He then checks the content against the national expert standard of the DNQP (the German network for quality development in nursing) and his facility’s own standard. Anything not covered there, he deletes.
Making technical texts understandable
An AI can translate technical terms into plain language, for example “pressure ulcer prevention” for an information sheet for relatives. Give it the general term or a passage from a brochure, never a doctor’s letter. Afterwards, read whether a fact slipped while it was simplifying.
Letters to relatives and ideas for the roster
Sara runs a home care service. From next month, the hours when the office can be reached by phone are changing. All relatives are to get a letter. Because the same letter goes to everyone, it needs no names, only placeholders.
Write a letter to the relatives of the people our home care service looks after.
Reason: from [date], our office can be reached by phone Monday to Friday from [time] to [time], no longer on Saturdays.
In urgent cases, the on-call service is still available at [phone number].
Tone: friendly, calm, formal. 150 words at most.
Thank them for their understanding without apologizing too much.
The greeting is “Dear relatives”. I am not adding any names.Before sending it, Sara checks whether the times and the number are right and whether the letter promises anything the service cannot deliver.
A roster without names
A real roster with names, leave requests and sick notes doesn’t go into a public AI chat. Sick notes are health data, including those of colleagues. What works: trying out rules and ideas with made-up codes.
I am planning shifts for a team of eight nurses, codes A to H, in a care home.
There are early, late and night shifts. Each early and late shift needs at least two qualified nurses.
Weekend requests should be shared out fairly.
Suggest three different rules for how we can grant requested days off fairly.
For each rule, name one advantage, one disadvantage and one question the team should settle first.
No statements on working time law or collective agreements; I will check those myself.Whether a suggestion fits working time law, the collective agreement and the works agreement is for the director of nursing to settle. The AI delivers ideas, not legal information.
What doesn’t belong in a public AI chat
- Diagnoses: an AI cannot examine symptoms. It can be wrong and sound just as sure. Assessments of someone’s condition are made by qualified staff; diagnoses are made by the doctor.
- Medication: doses, interactions and changes belong in the hands of a doctor or pharmacy. Even a general question about it doesn’t replace an order.
- Records with real data: care notes, wound documentation, life stories, vital signs. Not typed up, photographed or as a screenshot either.
- Photos of wounds or people: even without a face, a picture combined with date and place can make someone identifiable.
- Cases meant to be anonymous but with rare details: “resident, 94, formerly the village mayor” gives a person away even without a name.
Replacing a name with a code doesn’t make a case anonymous. Whoever has the list of codes knows who is meant. A case is only truly anonymous when nobody can identify the person any more, not even indirectly. With the people you care for, that is rarely certain. So leave such cases out altogether.
What your facility has to settle
Some facilities use AI services approved for their work. For this, the facility signs contracts with the provider and decides which data may go in. For data under professional secrecy, a data processing agreement (DPA; in German AVV) alone is usually not enough: the facility generally also has to bind the service to confidentiality and check where the data is processed. For hospitals and care services that bill health insurers, extra rules apply to cloud services. This is not something you decide at the ward computer, nor with your personal account.
You can ask the director of nursing or the data protection officer these questions:
- Do we have a rule on using AI?
- Which AI is approved, and which account do I sign in with?
- Which data may go in, and which never?
- May I use another service for templates and training without patient data?
- Whom do I contact if something was entered by mistake?
Since 2025, the EU AI Act has also required organizations that use AI to promote the AI literacy of their staff. A short rule and a briefing are a good start. If your workplace has neither yet, you can ask for them.
Check every result professionally
Even a template or training without patient data can be professionally wrong. An AI writes fluently even when a detail is out of date or made up. Before a text goes to the team or to relatives, check it:
- Does every professional statement match the expert standard, your facility’s standard or the doctor’s orders?
- Does the text give figures, deadlines or limits? Then look up each one in the source.
- Does it promise relatives anything you cannot deliver?
- Is really nobody identifiable, not even through details?
Delete anything that isn’t backed up. Professional responsibility stays with you and your team.
How to practice this
Writing instructions so that nobody is identifiable and the AI still delivers what you need takes practice. In the app “KI können” you practice this in short lessons, including examples from nursing, and put the instructions to work in your own AI. The trial lesson needs no account. How the app itself handles AI and data is explained under Notes on AI.
Sources
- General Data Protection Regulation, Art. 4(15) and Art. 9 (EUR-Lex) (opens in a new tab)
- German Criminal Code (Strafgesetzbuch), Section 203 Violation of private secrets (in German) (opens in a new tab)
- Regulation (EU) 2024/1689 on artificial intelligence, Art. 4 AI literacy (EUR-Lex) (opens in a new tab)
- German Data Protection Conference (DSK), guidance on AI and data protection (2024, in German) (opens in a new tab)
Information as of November 15, 2026. Laws and offers can change; this article is not legal advice.
Read more
Try it now
The trial lesson with examples for Registered nurse. No account, a few minutes.
Start the trial lesson