European digital sovereignty

Built in Europe. Your data stays here by default.

Rubrica runs in the EU by default: AI served from the EU, European hosting, a European database. Models outside the EU stay off unless your administrator enables them and the teacher accepts. So your university can adopt AI without worrying about where student data goes.

The European stack

Every layer of Rubrica was chosen to keep your data - and the AI that processes it - inside the EU.

AI models
Mistral AI · Paris 🇫🇷 · Quick AI · Quickers (coming soon) 🇪🇺 · OVHcloud · France 🇫🇷

By default, generation runs on Mistral Large, vision on Ministral and document embeddings on mistral-embed, served from the EU under API terms that exclude training on your data. When the institution enables them, a teacher can also pick: Quickers' own Quick AI and Quick AI Code (coming soon), served in a datacenter in the Netherlands; and open models served by OVHcloud in France - Qwen 3.5 9B, 3.6 27B, 3.8 27B and 3.5 397B, Qwen 2.5 VL 72B (with vision), Llama 3.3 70B, and gpt-oss 20B and 120B, OpenAI's open weights. Every one of them runs in the EU. Models outside the EU (Anthropic, OpenAI and DeepSeek, via OpenRouter) stay off unless your administrator turns them on and the teacher accepts.

Hosting & compute
European datacenters 🇪🇺

Self-hosted in Docker on European infrastructure. No US hyperscaler (no AWS / GCP / Azure), no Vercel. One stack, one jurisdiction.

Database · Auth · Storage
Self-hosted Supabase 🇪🇺

Postgres, authentication and file storage all run on our own EU server - a single European database you can export or self-host on premises at any time.

No AWS / GCP / Azure No Vercel Inference in the EU by default No data sold, ever

What this means for your institution

EU data residency

Every byte - accounts, quizzes, attempts, documents, reports - lives on European servers. Nothing is replicated to a non-EU region.

GDPR-first, DPA-ready

Lawful basis, consent capture, data minimisation, and the right to erasure are built in. We sign a Data Processing Agreement with your institution.

Aligned with the EU AI Act

Transparency about AI use, human oversight of every output, and no high-risk automated decisions about students. The teacher always decides.

Structural anonymity

Peer and jury ratings are aggregate-only by design - individual votes are never stored in a way the platform can expose. Privacy enforced in the schema, not just the UI.

AI for teaching, without the sovereignty trade-off.

Give your faculty a modern AI assistant and keep every guarantee your data-protection office needs.