What Is Proton Lumo 2.0? The Privacy-First AI Under Swiss Law

What Proton Lumo 2.0 is

Proton Lumo 2.0 is the new version of a privacy-focused AI assistant from Switzerland's Proton, known for email and VPN. Its defining trait is a design for people who want to avoid "having their conversations read in exchange for convenience," and this update also boosts performance significantly. For the wider trend of AI that stays self-contained within a jurisdiction, see our guide to sovereign AI (Apertus).

Swiss law, zero-access encryption, European infrastructure

Lumo 2.0's foundation is thorough privacy design. It is protected by Swiss privacy laws and zero-access encryption, and runs on Proton's European infrastructure. Zero-access encryption means that Proton itself, as the provider, cannot read the stored contents.

For work where you want to avoid having conversations used for training or passed to others, this "cannot be read" design is itself the reassurance.

More than 10 million users

Even with a sharply privacy-focused position, adoption is spreading. Proton says more than 10 million people have started using Lumo.

It's a figure that shows steady demand for AI you can use practically while keeping privacy intact.

A big jump in the intelligence score

Beyond privacy, the underlying ability improved a lot too. Let's check that gain with numbers from a third-party index.

From 15 to 51 on an independent index

The performance gain shows up clearly in a third-party metric. On the independent Artificial Analysis Intelligence Index, Lumo 2.0 Max scored 51 and Lumo 2.0 Lite scored 34. Since the old Lumo 1.4 scored 15, Max is 240% higher and Lite is 127% higher.

Because these are numbers on a shared external index rather than an in-house metric, the gain is more convincing.

Comparisons with other models are from reporting

Comparisons with other AI models should be read carefully. External media such as GIGAZINE report that Lumo 2.0 Max exceeds Claude Sonnet 4.6's score of 44, but this comparison does not appear on Proton's official blog. What the company states officially is only the gain between its own models (1.4 to 2.0). For the broader picture on Claude, see our guide to Claude.

New features and where to use it

Along with the performance boost, the available features widened too. Here's what it can now do.

Two reasoning modes, images, doubled context

The new features center on practical usability. It adds two reasoning modes you can pick by task — "Fast" (speed) and "Thinking" (for complex, multi-step reasoning) — plus image generation, editing, and analysis, and a doubled context length.

Use Fast for light questions and Thinking for work you want it to reason through — letting you trade speed and accuracy by situation.

Where privacy-first use fits

Lumo 2.0 fits work where you don't want the content to leave your hands. Contracts, internal documents, exchanges involving personal data — content you'd hesitate to hand to a general AI service — become easier to discuss with a "cannot be read" design. If you want to keep everything on your own hardware, our practical local-LLM coding guide is a useful comparison point.

In short, Lumo 2.0's biggest step is "reaching practical intelligence while keeping privacy intact." Privacy and performance tend to trade off, so showing that combination with external-index numbers is what makes it notable.

When passing material to an AI like Lumo, converting web pages to Markdown first keeps their structure intact and tends to improve reading accuracy.

Free ToolURL to Markdown ConverterConvert any public web page URL to Markdown. Preserves headings, tables, lists, and links — perfect for LLM and RAG preprocessing, research notes, and archiving web articles.Try it now →