
QELVIA Privacy Policy
Last updated: 5 August 2026 · Controller: Qelvia Solutions, S.L. (B88840251) · Contact: ouail@qelvia.ai
QELVIA is a conversational assistant with artificial intelligence that connects to your company's management system (ERP). This policy explains what data the application processes, how it obtains it, what it uses it for and who it shares it with.
1. What data we process and how we obtain it
• Messages you type to the assistant and files you attach (images, PDF) — you enter them yourself in the chat; they are used to generate the assistant's response.
• Business data from your ERP (stock, sales, invoicing, accounting…) — queried on your own company's server when your question requires it; used to answer with your company's real data.
• Your voice audio — only if you use voice dictation or read-aloud, with your microphone permission; used to transcribe your question or synthesise the response.
• Login credentials (username and password) — you enter them when signing in; the password is stored encrypted (hashed) on your company's server.
• Conversation history and assistant memory — generated through use; stored on your company's server, not in Qelvia's cloud.
2. Who it is shared with
To function, QELVIA sends data in encrypted form (HTTPS) to two AI providers, and only for the stated purpose:
• Anthropic (Claude language model): receives your messages, the files you attach and the ERP data needed to produce each response. Under its commercial terms, Anthropic does not use this data to train its models. (anthropic.com/legal/privacy)
• ElevenLabs (voice): receives your dictation audio to transcribe it and the response text to synthesise it, only if you enable the voice features. It does not use this data to train its models. (elevenlabs.io/privacy)
Both providers offer data protection guarantees equivalent to those in this policy. QELVIA does not share your data with any other third party, does not sell it and includes no advertising or trackers.
3. Local mode: no data leaves your company
Subject to prior agreement with the client, QELVIA can be deployed in local mode: the AI model runs entirely on the company's own servers. In this mode no data leaves the premises — zero data retention outside the organisation (ZDR) — and nothing is sent to Anthropic or ElevenLabs.
4. The app asks for your consent
The first time you open QELVIA, the application shows you this information and asks for your explicit consent before sending any data to the AI providers. If you do not accept, nothing is sent.
5. Where your data is stored
QELVIA follows a "your data in your house" architecture: the history, the assistant's memory and the credentials live on your own company's server (or on the managed server your company contracts), not on Qelvia's systems. The AI providers process each request's data to generate the response and do not retain it beyond their operational periods.
6. Device permissions
• Microphone: only for voice dictation.
• Camera and gallery: only to attach photos to your queries.
All permissions are optional and are requested at the moment you use each feature.
7. Your rights
You can exercise your rights of access, rectification, erasure, objection, restriction and portability (GDPR and LOPDGDD) by writing to ouail@qelvia.ai. You can also lodge a complaint with the Spanish Data Protection Agency (aepd.es).
8. Changes to this policy
If this policy changes, we will publish the new version at this same address with the updated date.