AI for clinics and healthcare: patient data stays in
- Updated
- By
- The MagnolIA team
- On this page
- 14 min read
Short answer
A clinic, mutual insurer or health insurer can use AI without sending health data to an outside provider. MagnolIA installs a machine in its network: documents containing patient data are read there, and the answers are produced there too. It helps with summarising and drafting; clinical decisions remain with the professional.
What is the problem with sending a medical record to a cloud AI?
Health data carries reinforced protection. Article 9 of the GDPR treats it as a special category. In Spain, Law 41/2002 on patient autonomy gives everyone the right to have the confidentiality of their health data respected (article 7) and binds staff who access the medical record to a duty of secrecy (article 16).
With a cloud AI, the report travels inside the prompt and is processed on another company's servers. Spain's Ministry of Health addresses this in its guidance for healthcare professionals on the use of AI (in Spanish): patients' personal data must not be entered into AI tools that are not designed, validated and authorised for healthcare use, or that offer no guarantees of GDPR compliance.
The same document considers it advisable to use general-purpose models as an auxiliary tool for summarising, drafting reports and other administrative, repetitive tasks, and advises against them for critical clinical decisions or as a substitute for professional judgement.
Real case: documents with patient data that never leave the office
The client is an insurance-sector company that works for healthcare professionals. We do not publish its name. The machine was installed in September 2026, and the route of the document is drawn in the on-premise AI use cases.
- The situation
- Their files are medical records, claims, complaints and expert reports: documents with health data. Sending them to a cloud AI meant handing them to a third party. They could only use AI if the document stayed inside the company.
- What was installed
- A MagnolIA machine in their own office, with MagnolIA Server and the language model loaded on that same machine. MagnolIA installed it, maintains it and updates the model.
- What they do today
- They open PDF, Word and Excel documents from their computers and ask MagnolIA to read them and draft from them. One example: a Word letter written from a PDF.
Why doesn't patient data leave? Five safeguards from the case
The model is in the office
It runs on the machine installed there. The document is read and queried on that machine.
It does not depend on any external service
Before handover it was tested with the internet cut off, and it kept reading and answering.
It is not connected to any cloud
Neither MagnolIA's nor anyone else's: it was delivered without being enrolled in any MagnolIA service.
The connection is encrypted
Between each computer and the machine, with a certificate issued in the installation itself.
Each person has their own user
They sign in with their own user and password, not a shared account.
What does AI do in a clinic, a mutual insurer or a health insurer?
| Use | Goes in | Comes out |
|---|---|---|
| Summarising and preparing reports | Clinical reports as PDF or Word files | A summary or a draft for the professional to review |
| Checking protocols | The centre's protocols and internal guidelines | An answer based on those documents |
| Administrative support | Clinical reports and the centre's paperwork | Letters, prepared appointments and support with coding |
| Handling a file | Medical records, claims or expert reports | A Word letter written from the PDF |
| Sessions and meetings | The audio of the meeting | Transcript and minutes with a summary |
| Sharing a document | A document with patient data | A copy without personal data and the key to revert it |
What changes at reception, in the consulting room and in the back office?
The table above sorts the uses by document. This one sorts them by role, with what each person keeps doing: in every case the AI hands over a draft and someone at the centre reviews it before it goes anywhere.
| Role | What the AI prepares | What the person still does |
|---|---|---|
| Reception and patient services | Draft letters and replies based on the centre's own documents; prepared appointments | Check the details, decide what is sent and through which channel |
| Clinical staff | The summary of a long report, a draft report, the answer to a question about an internal protocol | Check it against the original, correct it and sign. The clinical decision is theirs |
| Administration and billing | Support with coding, letters to insurers, reading PDF, Word and Excel files | Validate codes and amounts before relying on them |
| Management and quality | Minutes of internal sessions and meetings with a summary; questions put to the centre's guidelines | Approve the minutes and decide what is kept |
What should AI not do in a healthcare centre?
The guidance published by Spain's Ministry of Health separates acceptable uses from unacceptable ones. These five matter most to a clinic; the criterion comes from that document (quoted here in our translation) and the last column is the cautious alternative.
| Use | What the document says | What to do instead |
|---|---|---|
| Letting the AI diagnose or decide on a treatment | AI is a support tool and “under no circumstances may it be allowed to take autonomous decisions in healthcare” | Use it to read, summarise and draft; the decision belongs to the professional |
| Supporting a diagnosis with a tool that has no CE marking | “An AI without CE marking cannot be used to support diagnosis, treatment or any other function that influences direct patient care” | For that purpose, a medical device assessed for it. MagnolIA is not presented as one |
| Entering patient data into unauthorised tools | It must not be entered into tools that are not designed, validated and authorised for healthcare use, or that offer no data protection guarantees. Among unacceptable uses it also lists “requests that involve sending non-anonymised sensitive data to the AI” | Work with the tools the centre has authorised, or with anonymised information |
| Replying to patients automatically without telling them | It lists “automatic communications without warning the other party” among the unacceptable uses | Have a person review and send; if a system ever replies, say so |
| Sending identifiable results by WhatsApp or ordinary email | It asks for secure channels and rules out ordinary ones for identifiable clinical data and images | The corporate channels the centre has in place |
Is MagnolIA a medical device?
MagnolIA is not presented as a medical device. The definition in Regulation (EU) 2017/745 on medical devices (Article 2(1) of the text published in 2017) includes software that its manufacturer intends for specific medical purposes, such as the diagnosis, prevention, monitoring, prediction, prognosis, treatment or alleviation of disease. The definition turns on the purpose the manufacturer intends the product for.
MagnolIA is meant for office work: reading, summarising, searching and drafting. It advertises no CE marking as a medical device and is not meant to support a diagnosis. According to the Ministry's guidance, the absence of that certification does not make a tool unsuitable, but it limits its use to non-clinical purposes. That is the ground this page covers.
What if the AI gets it wrong? Who is responsible?
A language model can summarise a report badly, leave out a detail or write something the document does not say. That is why its output is a draft. The Ministry's guidance puts it this way: AI systems have no legal personality, and the professional must validate and supervise every recommendation before deciding.
On who answers for an adverse event, the same document gives no single rule: responsibility “does not fall automatically” on the professional, the organisation or the supplier, and is established case by case according to where the harm originated. Technology does not settle that question; it is a reason to write down who reviews what.
- Check against the original. A summary saves reading time; it does not replace opening the report when something depends on it.
- Keep drafts and signed documents apart. What the AI produces does not enter the medical record or leave the centre until someone has read it and taken ownership.
- Write the failures down. Where a failure is repeated, unusual or significant, the Ministry's document advises recording it and reporting it through the centre's internal incident procedure.
- Be able to reconstruct what happened. In a MagnolIA installation each person signs in with their own user, and the usage log and audit trail stay in the centre.
Does the patient have to be told?
The guidance distinguishes two cases. An informational use, with no personal data and no effect on care, does not require it. A care-related use, with a medical device or a tool that influences diagnosis or treatment, does. Elsewhere it asks for patients to be informed whenever AI plays a relevant part in the care process, including by “processing sensitive data”. Administrative tasks involving patient data are not settled by that summary: whether the centre's existing privacy information already covers them is for its data protection officer to assess.
What should you ask an AI provider before giving it health data?
The eight general questions (where it is processed, how long it is kept, who sees it, whether it is used for training…) are in what to ask an AI provider. With health data, add these five and ask for the answers in writing. They apply to any provider, MagnolIA included.
What use is the tool intended for?
General-purpose, or a CE-marked medical device. The answer determines which tasks the centre can use it for.
Who outside the centre could ever see a document?
Support, maintenance, other companies. In a MagnolIA installation the document is processed on the centre's machine and MagnolIA receives only the monitoring of the machine's status.
What data protection role does the provider take on, and under which contract?
A provider that maintains the machine is still an outside provider. How that relationship is documented is for your data protection officer to review.
Is there a record of who consulted what, and where is it kept?
It is worth being able to explain who accessed which documents. Ask whether the log is per person and whether it stays in your installation.
What happens to the data and the hardware at the end?
Deletion, return and deadlines. It is better to know before loading the first document.
Where does a clinic start? Steps and a checklist
List tasks, not tools
Which documents are read and written every week, and by whom. The candidates are repetitive reading and drafting tasks.
Talk to your data protection officer before choosing
With health data, they are the one who assesses the legal basis, whether a data protection impact assessment is needed and how patients are informed. The framework is in AI and the GDPR.
Start with administrative work
Letters, summaries, questions about protocols and minutes. The Ministry's guidance recommends general-purpose models as an auxiliary tool for finding information, drafting reports and other administrative, repetitive tasks.
Write down who reviews each output
One line per task: who asks, who reviews, and what is not done with AI. It is the basis of the centre's internal rules of use.
Run a pilot and measure
With a small group and the centre's real documents, inside its network. Note the errors found, not only the time saved.
Checklist before the first real document
- We know which machine processes the document and who maintains it.
- Each person has their own user; there are no shared accounts.
- The data protection officer has seen the intended uses.
- It is written down which tasks are not done with AI: diagnosis, treatment and automatic communication with patients.
- Every type of output has a named reviewer.
- There is a place to record the AI's errors.
- Nobody uses personal accounts on public AI services with patient data.
How is it installed, and what stays inside the centre's network?
MagnolIA delivers it installed and keeps it running: the centre does not administer servers or models. The reference deployment for a clinic starts on Núcleo M, and capacity is added without interrupting service.
Assessment
We look at how many people will use the AI, what for, and which systems it needs to talk to. That sets the capacity.
Installation
MagnolIA installs the machine in your office. You do not buy it: it is included in the fee.
Tuning and integration
The models are loaded and tuned, and the integrations with your systems are set up.
Pilot
A group of 10 to 20 people works with it on their own documents.
Production
The whole staff, with corporate sign-in, monitoring and backups.
- Stays inside
- Prompts, documents and answers. They are processed on the machine installed in your network and stay there.
- Also inside
- The per-user usage log and a chained, verifiable audit trail, kept for two years.
- Leaves
- Only the monitoring of the machine's status: load, health and security. MagnolIA does not see the content.
- Access
- Each person signs in with their own user and password through a single entry point encrypted with TLS. Remote access is optional and runs through an authenticated outbound tunnel.
The measures are in on-premise AI security, the payment model in what on-premise AI costs and the basic definition in what on-premise AI is.
Can you try it before installing anything?
Yes. The online document anonymiser returns the document with names, ID numbers, addresses and accounts replaced, plus the key to revert it, and the personal data detector flags what a text carries before you paste it into a public AI.
Frequently asked questions from clinics and healthcare
Can AI be used with patient health data?
Health data is a special category under Article 9 of the GDPR, so using AI on it means controlling where it is processed. For a client in the insurance sector that works for healthcare professionals, MagnolIA installed a machine in its office: documents containing patient data are read there and the answers are produced there, with no reliance on any external service.
How can a clinic use AI with medical records?
A clinic can use AI on medical records by keeping the data inside its own network. Medical records are special-category data. With a MagnolIA machine installed there, clinical and administrative staff get summaries and draft reports, answers about internal protocols and help with appointments, letters and coding, without the document leaving. Which tool is authorised, and for which uses, is for each centre to decide with its data protection officer.
Can AI diagnose or decide on a treatment in a clinic?
According to the guidance for healthcare professionals published by Spain's Ministry of Health, AI is a support tool and may not make autonomous decisions in healthcare: the final word and the responsibility lie with the professional. A general-purpose AI such as MagnolIA is for reading, summarising and drafting documents that someone at the centre then reviews.
Who is responsible if AI gets it wrong in a healthcare centre?
When AI gets something wrong in a healthcare centre, responsibility does not fall automatically on the professional, the organisation or the supplier: according to guidance from Spain's Ministry of Health, it is established case by case from the origin of the harm. AI systems have no legal personality, so the professional validates and supervises every result before deciding.
Is MagnolIA a medical device?
MagnolIA is not presented as a medical device and advertises no CE marking for that purpose. Regulation (EU) 2017/745 puts in that category software its manufacturer intends for medical purposes such as diagnosis or treatment. MagnolIA is meant for office work: reading, summarising and drafting, not for supporting clinical decisions. Each centre assesses where a given use falls.
What is the difference between anonymising and pseudonymising?
Anonymising means removing the link to the person irreversibly. Pseudonymising means replacing the data and keeping the information needed to restore it separately. In GDPR terms (Article 4(5)), what MagnolIA Nexus does is pseudonymise: the mapping table stays on your machine and the AI never receives it.
Sources
External figures on this page link to their original source.
EUR-Lex. Regulation (EU) 2016/679, General Data Protection Regulation (GDPR). eur-lex.europa.eu/eli/reg/2016/679/oj
BOE (Spain's Official State Gazette). Law 41/2002 on patient autonomy and clinical records, articles 7 and 16 (in Spanish). www.boe.es/buscar/act.php?id=BOE-A-2002-22188
Spanish Ministry of Health. Guidance for healthcare professionals on the use of AI, published under the AI Strategy for the National Health System (PDF, in Spanish). www.sanidad.gob.es/areas/saludDigital/estrategiaIASNS/doc/IASNS_Orientaciones_para_profesionales_sanitarios_en_el_uso_de_la_IA.pdf
Official Journal of the European Union. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices (Article 2, definitions; text published on 5 May 2017, Spanish edition, PDF). www.boe.es/doue/2017/117/L00001-00175.pdf
Shall we look at it with your centre's documents?
Tell us which documents you work with and how many people would use it. The pilot runs with a group of 10 to 20 people.
Request a pilot


