AI transparency
Last updated: 2026-10-04
Owivo is the AI receptionist of a dental practice. A patient asks a question in the chat window on the practice website, and an AI language model writes the answer from the information the practice entered itself. This page sets out which models those are, what is sent to them and what happens to it.
Every answer takes one route: Owivo sends the question to OpenRouter, which passes it to a Claude model on a European endpoint. There is no other provider as a backup.
Forwards the question to the model
OpenRouter is a US company that forwards the request to the model endpoint, so the request passes through OpenRouter. Owivo sends every request with four settings: no data collection, only zero retention endpoints that do not keep the content and do not train models on it, only the European endpoints Amazon Bedrock eu-west-1 and Google Vertex AI europe, and no permission to fall back to any other endpoint.
The language model that writes the answer
Answers to patients come from Claude Sonnet 5. Internal jobs, such as judging how hard a question is, run on Claude Haiku 4.5. Both models run only on the two European endpoints above. When both are unreachable, Owivo does not fall back to an endpoint outside the EU.
All traffic between the patient browser, Owivo and the providers runs over TLS. API keys are stored as environment variables on the server and are never loaded into a patient browser. Each practice sees only its own data.
A question about how Owivo uses AI? Send it to: info@owivo.com