Discharge letters in 20 instead of 45 minutes. 50 pages of patient history understood in 5 minutes. Medical confidentiality preserved.
anymize gives physicians and hospitals access to the world's best AI models – without breaching the duty of medical confidentiality or ignoring GDPR Art. 9. The anonymization strips names, dates of birth, diagnoses and case identifiers from every document before it reaches a frontier model.
The promise
The physician who knows 50 pages of patient history in 5 minutes. The senior physician who has time for the conversation because the discharge letter almost writes itself. The doctor who burns for medicine – not for documentation.
You became a physician because diagnoses, patient conversations and treatment journeys fascinate you. The reality: a third of your working time goes into documentation – discharge letters, findings reports, insurer correspondence, quality indicators. Elsewhere, AI has long taken over this routine work. In your field it fails on data protection and the duty of medical confidentiality.
anymize opens the door. Patient documents are anonymized before any AI access; the answer returns with real names. From the AI provider's perspective, they see anonymous cases. From your perspective, you have a structured discharge letter based on your real patient history – in a fraction of the time.
Compliance framework
The hard compliance boundaries.
Patient health information is special category data under GDPR Art. 9, requiring explicit legal basis for processing. Healthcare providers must ensure that any AI tool handling patient data meets the heightened requirements for sensitive data — including strict purpose limitation, security measures, and contractual safeguards with every processor.
Healthcare professionals carry a universal duty of medical confidentiality — binding across all jurisdictions from the EU to the UK, US, and beyond. This duty extends to AI tools: any system that can access patient data must provide equivalent confidentiality guarantees. Anonymization before AI processing ensures the tool never 'sees' protected health information.
anymize provides by default:
Under the EU NIS2 Directive (effective since October 2024), hospitals and healthcare providers are classified as essential entities subject to mandatory cybersecurity requirements. This includes supply chain security, incident reporting, and risk management — all of which apply to AI tools integrated into clinical workflows.
The EU AI Act classifies AI systems used for medical diagnosis, treatment decisions, and patient management as high-risk AI. Healthcare providers deploying AI tools must maintain conformity documentation, audit trails, and human oversight mechanisms. anymize's anonymization layer creates the evidentiary record needed for AI Act compliance.
Medical documentation requirements apply across jurisdictions. AI systems support documentation but do not replace the physician's review and responsibility.
Digital health transformation programs across the EU and beyond support hospital modernization. anymize, as an AI tool, can be deployed in such initiatives; we support eligibility questions with technical evidence on request.
Under NIS2 critical entity obligations, larger hospitals must meet elevated cybersecurity and resilience standards. On-premise variants reduce cloud-disclosure load and simplify demonstrating compliance.
Five everyday scenarios
Five typical healthcare workflows – and what they look like with anymize.
01 · Scenario
Tool chain
Project "Ward 3A" active · knowledge base "Hospital standard letters" · Claude Deep · upload findings (anonymized) · "Draft a discharge letter following the hospital template"
Outcome
Draft as an artifact, editable in WYSIWYG, original export for the patient record, anonymous export for peer review
02 · Scenario
Tool chain
Live transcription (smartphone at the bedside via QR) · speaker diarization for physician/patient/nurse · transcript as document · Fountain: "Structure as a ward-round note with medication changes"
Outcome
Finished ward-round note + medication reconciliation
03 · Scenario
Tool chain
Upload lab findings (anonymized) · Claude Smart · "Summarize for a general practitioner, 200 words, no jargon"
Outcome
Patient-friendly letter as an artifact
04 · Scenario
Tool chain
Perplexity · "Current guideline recommendations for indication X with comorbidity Y" · cross-reference to internal knowledge base "Hospital SOPs"
Outcome
Evidence-based summary with source citations
05 · Scenario
Tool chain
anymize Fountain (EU-hosted inference, no anonymization required for internal forms) · upload the form · "Check against diagnosis codes and billing rules, flag gaps"
Outcome
Structured answer, open questions flagged
Model recommendation
| Task | Model | Why |
|---|---|---|
| Discharge letters, physician letters, expert opinions | Claude (Deep) | Natural medical prose, professional register, precise wording |
| Clinical literature research | Perplexity | Web-native research with source citations for current guidelines |
| Multimodal findings (ECG, radiology sketches) | Gemini | Multimodal, image understanding |
| Very long patient records (oncology, chronic cases) | Kimi (Deep) | Very large context windows, one pass instead of chunking |
| Internal billing, insurer and statistics data | anymize Fountain | EU-hosted inference, no anonymization required |
| Research data analysis, complex reasoning chains | anymize Waterfall or Claude Deep | Waterfall for data-in-EU, Claude for frontier analysis |
Natural medical prose, professional register, precise wording
Web-native research with source citations for current guidelines
Multimodal, image understanding
Very large context windows, one pass instead of chunking
EU-hosted inference, no anonymization required
Waterfall for data-in-EU, Claude for frontier analysis
Feature set for hospitals and practices
The number one lever in healthcare.
Ward rounds, history-taking, tumor boards, case conferences – record via smartphone, transcribed on our self-hosted infrastructure in Germany, speakers automatically separated (physician, patient, nurse, senior physician). The transcript becomes a document the AI turns into discharge letters, notes, to-do lists. 13 languages including German.
One database per ward, department or client project – hospital-internal SOPs, treatment guidelines, coding manuals, team-specific templates. With person-to-person sharing, hospital physicians can share knowledge with cooperating practices.
One project per ward, specialty program (e.g. oncology, cardiology) or study. Context, instructions and linked knowledge bases are shared across all team members – individual chats remain private.
Discharge letters, findings summaries and patient information sheets are created as WYSIWYG-editable artifacts. Anonymous / original export toggle: peer review with colleagues anonymous, the final version for the patient record with original names.
Enforce a hospital-wide anonymization policy, four-eyes review before sending to frontier models, audit log for every processing event – indispensable in larger institutions with many roles.
Integration
Integration with existing hospital software – KIS, PVS, DMS and FHIR.
Integrations with ORBIS, medico, iMedOne, ISH-med and others via the API layer or automation platforms.
Connections to ifap, T2med, Medistar, x.isynet via workflow tools.
Direct access via the OpenAI-compatible anymize API.
Structured patient records can be processed through the anonymization pipeline.
Direct out-of-the-box integrations into the German KIS/PVS ecosystem are expanding; as an immediate path, hospitals use the API with n8n or Make.com.
On-premise relevance
For hospitals classified as critical entities under the EU NIS2 Directive, for university hospitals with research infrastructure, or for practices with particularly confidential patient groups (addiction medicine, psychiatry, HIV specialization), we offer on-premise deployments: anonymization plus our small model anymize Spring within your own infrastructure. Hybrid variants allow you to use international frontier models for top-tier answer quality, while all sensitive content stays inside the hospital rack.
For most practices and smaller institutions, our cloud solution is sufficient – the Data Processing Agreement under GDPR Art. 28 and our medical confidentiality commitment cover it.
Think hybrid.
Spring + frontier: what must stay local, stays local; what needs the best answer goes out anonymized – to GPT, Claude or Gemini.
Which plan fits
Start in three steps
The Data Processing Agreement under GDPR Art. 28 takes effect automatically on acceptance of the terms. For professionals bound by medical confidentiality (physicians, dentists, pharmacists), the onboarding additionally presents a medical confidentiality commitment – text form is sufficient, one click.
Start with your discharge-letter template plus 3–5 example letters (anonymized). The AI learns your style, your structure, your characteristic phrasing. After the first week, the generated draft is clearly closer to your institution's style than a cold-start draft.
Set admin policies (anonymization enforcement, audit log, optional review countdown). Hold a 30-minute demo in the ward meeting. In the first month, typically 3–5 use cases that the team identifies on their own.
Healthcare roles
Faster patient communication, physician letters from lab findings, form assistance
Guideline research, second-opinion support, responding to insurer inquiries
Faster discharge letters, ward-round follow-up, case-conference minutes
Enforce team standards, quality consistency, strategic analyses (Waterfall)
Literature reviews, cohort analyses, scientific writing assistance
Cross-site knowledge bases, uniform documentation, billing quality
Frequently asked questions
Not directly – unencrypted transmission of patient data to an external AI provider would breach your duty of medical confidentiality and GDPR obligations on special category data. Via anymize: yes. The anonymization removes identifying data before the document leaves your building. For the external model, no protected health information remains. We formalize this with a medical confidentiality commitment and a Data Processing Agreement under GDPR Art. 28.
Legal notice
The statements on this page regarding GDPR Art. 9, the duty of medical confidentiality, the EU NIS2 Directive and the EU AI Act constitute a classification based on the currently applicable regulations (as of April 2026). Regional and national implementing rules, as well as professional codes for physicians, may deviate. These statements do not replace individual legal or data-protection advice. For binding assessments, please consult your data protection officer, a specialist lawyer for medical law, or the competent medical regulatory body.
We stand behind anymize. And we know – when an AI tool touches client, patient or employee data, a demo video isn't enough. That's why we give you 14 days of full access – all models, all features, no credit card. Enough time to be certain, before you trust us.
Your AI workplace awaits.