Proton's Lumo assistant now runs on Apertus, Switzerland's sovereign open AI model
Proton has wired Apertus — a fully open large language model built by EPFL, ETH Zurich and Switzerland's national supercomputing centre — into its Lumo assistant, so a mass-market European product now runs a model developed and governed in Europe rather than an American one.
On 17 September 2026, Swiss privacy company Proton added Apertus, a fully open large language model built in Switzerland, to its Lumo AI assistant. Apertus was developed by EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS) — a publicly funded model whose weights and training documentation are published openly.
Inside Lumo, the model runs under the assistant’s existing privacy terms. Conversations stay protected by zero-access encryption, so Proton cannot read them and keeps no logs. Users can rate answers with a thumbs up or down, and that feedback is anonymised before it reaches the Apertus researchers — giving the open model a real-world improvement loop without exposing who asked what.
“We don’t want to spend the next decade relying on other countries for the AI we use every day,” said Eamonn Maguire, Proton’s director of AI. He framed the release as Europe pairing infrastructure it controls with a model that is open by design and a product that puts it directly in people’s hands.
Proton already appears here as the maker of Proton Mail and Lumo, its Swiss answers to Gmail and ChatGPT. Most tools marketed as “European AI” still call an American model underneath; wiring a European sovereign model into a shipping consumer assistant is a step past that — Lumo now runs a model built and governed in Europe, not an EU-hosted wrapper around a US one.
Apertus 1.5 is the version Proton is using. It is one of the first openly developed European models to power a mainstream consumer product rather than a research demo.
Sources: Proton (proton.me/blog/lumo-apertus-partnership, 17 Sep 2026); ETH Zurich news (17 Sep 2026).