Lumo 2.0: The most powerful private AI

My notes

Summary

Proton launched Lumo 2.0, a rebuilt version of its zero-access encrypted AI assistant that never logs conversations or trains on user data. It adds reasoning/fast modes, multimodal image handling, live web search with citations, user-controlled memory, and a Business tier, all on European infrastructure under Swiss privacy law. Positioned as the privacy alternative to US-hosted assistants that mine conversations for training and ads.

Key Insight

• Benchmark jump (self-reported): On the Artificial Analysis Intelligence Index, Lumo 2.0 Lite scores 127% higher than 1.4, and 2.0 Max scores 240% higher. Everyday queries respond up to 76% faster.
• Two model modes: Fast (speed) and Thinking (multi-step reasoning with a visible thinking state), a capability that did not exist in 1.4.
• Zero-access encryption extends to images and memory: uploaded/generated images and stored context are encrypted so not even Proton can read them. Same architecture securing 100M+ Proton Mail/Drive users.
• Now multimodal: image analysis (charts, screenshots, docs), generation, editing, and sketch-to-image, all in one conversation. Context window doubled.
• Business angle is the sharp differentiator: data stays on independent European infrastructure, so access “cannot be subject to US Executive Orders” or American data collection requests, a direct pitch to orgs worried about US-law exposure and vanishing access for international users.
• Fully open source: anyone can inspect the encryption code.
• Free tier exists; Lumo Plus (unlimited chats, Projects, advanced image gen) and Lumo Professional (team collaboration) are paid.