Lumo 2.0

Proton’s new AI gets faster and smarter, but the comments instantly turned into a privacy brawl

TLDR: Lumo 2.0 adds faster answers, image tools, memory, and stronger web search while selling itself as a privacy-first AI. Commenters love the European privacy angle, but the thread quickly turned into a fight over whether Proton is really as “open” as it sounds.

Lumo 2.0 has arrived promising a big glow-up: faster replies, a new deeper-thinking mode, image creation and editing, web search with sources, longer memory, and private workspaces for ongoing projects. Proton is pitching it as the AI for people who want help without handing over their secrets, loudly reminding everyone that it runs on European servers under Swiss privacy laws. For some commenters, that was the whole selling point. One user basically said, "finally, an AI I might be allowed to show to clients who panic at the idea of American tools touching their code." Another was perfectly happy to pay for what they see as a rare thing in AI land: privacy plus decent conversation management.

But of course, the real action was in the backlash. The hottest mini-scandal? Open-source drama. Multiple commenters were not impressed by Proton’s branding, arguing that calling Lumo “open” feels slippery when, as they see it, only the app around it is open and the actual AI brain is not. One user even dropped a link like a receipt in a courtroom drama. Another said they trust Lumo only marginally more than ChatGPT, which is basically the internet equivalent of a raised eyebrow.

There was also some accidental comedy gold: a shoutout to Mistral’s coding tool, “Vibe-cli”, got the kind of reaction that makes you double-check it’s not a parody name. So yes, Lumo 2.0 launched with flashy new features — but the comments made it clear the real fight is over whether “private” and “open” actually mean what users think they mean.

Key Points

  • Proton launched Lumo 2.0 as a major rebuild of its encrypted AI assistant, maintaining zero-access encryption and European infrastructure protected by Swiss privacy laws.
  • The article says more than 10 million people have used Lumo since launch, and Lumo 2.0 significantly improves benchmark performance over Lumo 1.4.
  • Lumo 2.0 adds Fast and Thinking modes, with everyday responses up to 76% faster and a visible thinking state for complex tasks.
  • The assistant is now multimodal, supporting image generation, image editing, image analysis, and sketch-to-image workflows with encrypted image storage.
  • Lumo 2.0 also expands web search, user-controlled memory, larger context windows, encrypted Projects, and task-specific Custom Lumos.

Hottest takes

"don’t want US LLMs anywhere near their IP (code)" — teekert
"only their clients are open. The models themselves? Well, good luck with that..." — Signez
"I trust Lumo only marginally more than ChatGPT" — illright
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.