Italian healthcare and life sciences professionals face an impossible dilemma every day: spend hours manually searching for clinical protocols scattered across shared drives, or upload sensitive health data to a cloud AI tool and risk violating GDPR, the Italian AI Law L.132/2025, and the trust of every patient under their care. Noot 8 eliminates the dilemma.
The Status Quo Is Not Working
Imagine a clinical coordinator in a private oncology clinic. They need to cross-reference the current treatment protocol with a supplier agreement before a consultation starts in 20 minutes. They open the shared network folder. Three subfolders. Four versions of the same document. No context-aware search. They spend 15 minutes searching — and are still not sure they have found the latest revision.
This scenario repeats dozens of times a week, in every healthcare facility in Italy. The instinctive solution — pasting the document into ChatGPT or Microsoft Copilot — creates a compliance crisis:
- Article 9 of the GDPR classifies health data as a special category that requires protections that cloud processors hosted in the United States cannot structurally guarantee. The EDPB report of 2025 confirmed that LLM outputs rarely qualify as anonymised data.
- Italian AI Law L.132/2025, enacted in October 2025 as the first national AI law in the EU, imposes strict data localisation and transparency requirements for AI systems processing health information. The compliance deadline of August 2026 is already looming — and enforcement is not theoretical.
- Clinical governance policies in most Italian facilities, they explicitly prohibit the uploading of protocol documents or patient-related data to external services. Cloud AI is structurally incompatible with these obligations.
The most secure teams simply do nothing. They continue to search manually. They continue to waste time that belongs to patients.
Noot 8: The AI that never leaves your building
Noot 8 is a plug-and-play edge AI device built for organisations that handle sensitive documents and cannot accept exposure to the cloud. Plug it into your LAN. In less than five minutes, any authorised team member can search the entire library of documents — clinical protocols, patient pathways, research articles, regulatory submissions, vendor agreements — using simple natural language and receive page-level cited answers.
No cloud account. No IT setup. No user fees. No data leaves your network.
Eye Cortex: Visual Intelligence Built for Clinical Documents
Health documents are not simple text files. They are dense, multi-format artifacts: scanned consent forms, drug interaction tables, protocol flowcharts, regulatory PDFs with embedded annotations, research articles with complex mixed layouts. Standard AI for text extraction fails on all of these.
The Eye Cortex visual engine of Noot 8 completely changes this. Powered by a computer vision model with 4 billion parameters built on the ColPali architecture, Eye Cortex reads every page of a document as an image — just as a clinician does. It understands:
- Scanned PDFs and overlays of handwritten annotations
- Drug dosage tables, flowcharts of clinical protocols and diagnostic grids
- Regulatory submissions in mixed layout and published research articles
- Agreements with suppliers with embedded specification tables and pricing grids
Ask Noot 8: “What is the recommended second-line therapy in our 2025 oncology protocol?” It returns the exact answer with the source document and page number cited — in under 30 seconds — without a single byte of that protocol leaving your hospital's network.
The Privacy Guarantee Is Architecture, Not Policy
Most AI providers offer a privacy policy. Noot 8 offers a privacy architecture. For healthcare, the difference is crucial.
Here’s exactly what happens with every query on Noot 8:
- 100% local processing: All incorporation, indexing, and archiving of documents occurs on the Noot 8 device, on your LAN, within your facility. Your clinical documents never touch an external server.
- Only the retrieved extracts leave the device — never the source files: For complex reasoning, only the relevant extracted text fragments are sent to Google Gemini via Vertex AI, processed within the EU under Google’s Business Data Processing Agreement, with zero storage and zero model training on your data.
- Source documents always remain on-site: Your protocols, the documents on patient pathways and research archives remain on the device. This is not a configuration option. It is the architecture.
This design is structurally compliant with the data minimisation principle of the GDPR, the data localisation requirements of L.132/2025, and the provisions of the EU AI Act for high-risk AI systems in healthcare contexts — without requiring your legal team to check the terms of service of a cloud provider every six months.
This is not just a privacy policy — your clinical documents are guaranteed to remain on your hardware.
What Changes from Day One
- Clinical coordinators find the correct version of the protocol in seconds, not minutes
- Medical directors instantly cross-reference agreements with suppliers — without sending a document outside the building
- Compliance officers confirm that no patient-related data has ever been outsourced from the AI system
- Research teams search through hundreds of articles and internal study archives with page-level citation accuracy
All of this is available to every team member on your LAN simultaneously, under a single all-inclusive subscription at €149 per month. Hardware included. No installation costs. 30-day money-back guarantee.
Book your live demo today
Watch Noot 8 search your documents — clinical protocol, research PDF, regulatory submission — in real-time. No slides. No theory. Just your documents, answered in under 30 seconds, with the cited source page, without leaving your building.
The compliance deadline of August 2026 for L.132/2025 and the EU AI Act is not theoretical. Italian healthcare organisations that have not audited the use of their AI tools are already exposed. Noot 8 eliminates that exposure by design.