Medical practices & walk-in clinics
An AI service that has read your protocols and templates, hosted in Switzerland. Without your patients' data leaving Switzerland.
Consumer AI is not compatible with medical confidentiality
Some of your colleagues use ChatGPT. You can't.
Medical confidentiality (article 321 of the Swiss Criminal Code) does not allow patient data to be sent to a foreign server. ChatGPT, Gemini, Copilot — every consumer AI transfers your data out of Switzerland. A single dictated report containing a patient name is enough to create a potential breach.
Your evenings keep stretching writing reports and assessments.
Consultation reports, letters to colleagues, insurance reports (AI, AVS, RC), searching a patient's record, summarising medical history — all tasks that eat into your evenings instead of being delegated. AI could free you up, but consumer AI is simply not an option.
An AI service built for medical confidentiality
Your documents stay in Switzerland, under your control. Chat to try; the assistant for daily work. Many international vendors cannot offer you that.
Level II — Your protocols and templates
The AI has read your report templates, your protocols, your vocabulary. It searches, cites its sources; you review. This is the level we recommend.
Apertus, hosted in Switzerland
Swiss model (EPFL + ETH Zurich + CSCS), run at 2itea in French-speaking Switzerland. Compatible with article 321 of the Swiss Criminal Code and the nLPD. No prompt leaves Switzerland.
Four service levels
I Chat to try, available now. II Assistant, available now. III Agents after capacity is commissioned (6 to 12 weeks). IV Dedicated, on quote.
On your premises: level IV only
The service runs at 2itea. An installation on your premises only happens at the Dedicated level, on request.
The interface, at a glance
Accessible from a browser, like ChatGPT — but dedicated to your practice. At level II, the AI relies on your protocols and cites its sources. Apertus, hosted in French-speaking Switzerland.
Illustrative preview — conversations and suggestions are fictitious, for demonstration purposes.
A medical practice — typical case
A fictional example, to show what level II looks like in a Romandy practice. This is not a client in production.
The context
Primary-care medical practice, four physicians and their team of medical assistants. Weekly volume of consultations, reports to write, letters to specialists, insurance reports. Several years of archived patient records and protocols specific to the practice.
The deployment
Level II: the AI has read the report templates, care protocols and vocabulary of the practice. It searches, cites its sources; the practitioner reviews. Hosted at 2itea in French-speaking Switzerland — secure web interface, individual access. Audit trail of every use.
The expected outcome
Consultation reports drafted in a few minutes. Immediate search within the patient record. Faster preparation of insurance reports. And above all: no record ever leaves the practice.
Compatible with your legal obligations
The safeguards that make the platform legally usable by medical professions bound by medical confidentiality.
Medical confidentiality is protected under criminal law for physicians, nurses and their auxiliaries. The platform upholds the absolute confidentiality of your patient records — no data leaves your practice.
Federal Act on University Medical Professions. The platform is designed to fit within the physician's ethical and liability framework.
New Swiss data protection act (1 Sept. 2023). Patient data 100% in Switzerland on our infrastructure in French-speaking Switzerland. No dependency on a foreign cloud.
Tier IV datacenter in French-speaking Switzerland, operated by 2itea. No foreign provider in the platform's production chain.
Built on Swiss public academic research (EPFL + ETH Zurich + Swiss National Supercomputing Centre). Open, auditable model code and weights.
Every interaction with the assistant is logged and timestamped. You know who accessed what, when, for which patient. Compliant with traceability requirements.
Five steps to move forward together
From the first demo to ongoing operations — the same path for monitoring and artificial intelligence.