What we process
- Questions sent to the widget or API are processed to produce an answer from the relevant tenant's content. We do not attach visitor identities to questions — the widget sets no cookies and does no fingerprinting or cross-site tracking.
- Customer content (the documents a tenant uploads) is stored to serve that tenant's answers, encrypted in offsite backups, and never used for anything else — not for training, not for other tenants, not for us.
- Privacy-minimized operational records never store the words in a visitor's question. For up to 30 days by default, we retain a keyed question fingerprint, character count, language, answer outcome, evidence identifiers, citation hash, decision-receipt identifier, and latency so a tenant can measure coverage and investigate failures without reading visitor text. Raw-question logging cannot be enabled.
- Service and security logs (health, errors, and events such as failed admin logins) are kept to run and protect the service.
What never happens
- No third-party AI provider ever sees your content or your visitors' questions — the models are ours and run on infrastructure we or you operate.
- No advertising, no data sales, no trackers on this site.
- On-premise appliances process everything locally; nothing about your visitors reaches us at all.
Viavitna is operated by ASL LLC. Questions, or a data request? Contact us — a person reads it.