Who makes Evanesca, and why.
Evanesca is published by FLEECE AI, a French simplified joint-stock company created on 26 Aug 2024 and based in Chamalières. Its founder, Loïc Jané, is the publication director. Evanesca starts from a simple idea: you should be able to use AI for what matters to you, without leaving a copy of it behind.
The company
- Company
- FLEECE AI
- Legal form
- Simplified joint-stock company with a single shareholder (SASU), share capital €1,000
- Created
- 26 Aug 2024
- Registered office
- 10 rue Pierre Poisson, 63400 Chamalières, France
- Registration
- 932 329 188 R.C.S. Clermont-Ferrand — SIRET 932 329 188 00019
- EU VAT number
- FR45932329188
- Founder
- Loïc Jané, publication director
- Contact
- contact@fleeceai.agency · +33 6 06 77 70 55
The same details are in the legal notice.
Why Evanesca exists
Most consumer AI chats keep your conversations until you delete them, may train their models on them unless you opt out, and are run by US companies that the CLOUD Act can reach — their own policies say so, quoted on our sources page. That is why more than one company in four has banned generative AI at some point (Cisco, 2024): what you would most like to ask is often what you should not send.
Evanesca takes the other road: a private AI agent and chat whose models all run in Europe — in France, in Mistral AI's EU region, in Germany or in Switzerland — through European operators; the sovereignty page says who controls each. Your messages never leave Europe; Evanesca keeps no copy of them, and your history — with the agent's files — stays in your browser; it charges for what you use, with no subscription; and under every answer it shows a passport saying where the answer was produced and what was kept.
What we hold ourselves to
Say where, and who
Every answer names its model, its operator and its country. Every model runs in Europe, and no model call leaves it — not as a backup, not when servers are busy.
Keep no copy
Your messages stay in memory while the answer is written, then Evanesca forgets them; about your messages, we keep only billing data. Two operators keep more: Mistral AI keeps the input and output of its API 30 rolling days to monitor abuse, and Scaleway may keep requests that cause errors or look abusive for up to 14 days — the passport says so.
Test before trusting
Every model shows its six checks before you use it, and a missing stamp stays visible.
Write only what can be proved
No invented figures, no badges we have not earned: every fact on the site is computed from the catalogue or linked to its source, with the date it was read.
Where Evanesca stands
The chat is not open yet: this site shows how it will work. The host, the database and the payment and email providers are named in the legal notice and the privacy policy; before it opens, we will also name the consumer mediator in the terms.