AI Safety & Guidelines
Learn about how Klinio AI operates as a secure clinician-supervised drafting assistant, and our strict rules on clinical safety.
Klinio AI is a documentation drafting assistant. It does NOT diagnose diseases, does NOT make treatment decisions, and does NOT replace the clinical judgment of a licensed healthcare professional. Every AI-generated output is a raw draft and must be reviewed, edited, and approved by the clinician before use.
1. Intended Use and Capabilities
Klinio AI is designed to reduce the administrative burden on clinicians by converting clinical observations and notes into structured text drafts. Safe use cases include:
- Patient Summaries: Formatting daily treatment inputs into readable patient chart summaries.
- Proposal Drafts: Preparing estimates, billing lists, and invoice drafts based on clinic price lists.
- Patient Communications: Drafting appointment reminders, follow-up advice (e.g. after-care rules), and WhatsApp message templates.
- Document Drafts: Generating drafts of letters or administrative texts.
2. Unacceptable and Dangerous Use Cases
Clinicians must never use Klinio AI to:
- Diagnose patient symptoms or evaluate laboratory results.
- Define prescriptions, drug dosages, or medical therapies.
- Determine radiological evaluations or disease presence.
- Replace professional consultation or act as a standalone clinical recommendation system.
3. Clinical Responsibility
The licensed professional retains sole clinical responsibility for patient health, prescriptions, clinical documentation, and care parameters. Klinio operates only as an administrative text formatting utility. Klinio is not intended to operate as regulated clinical decision software.
4. AI Privacy and Security Controls
We process AI prompts over private, secure endpoints. To maintain patient privacy:
- Name Sanitization: The application is configured to filter out direct patient identifiers (such as national ID numbers or exact addresses) before sending prompts.
- No Model Training: Prompt contents and clinician feedback metrics are not used to train public language models.