Proton's Lumo 2.0 adds image tools, reasoning and long-term memory

Proton rolled out Lumo 2.0 with image generation, image recognition and editing, advanced reasoning, deep private web search and long-term memory, in its biggest upgrade yet to the privacy-first assistant. The company is using the release to make a broader case that users do not need to trade privacy for capability in generative AI, even as OpenAI and Anthropic keep pushing the technical ceiling higher.

Proton says Lumo is protected by zero-access encryption, keeps no logs and does not use conversations to train its language models. The assistant is built on open-source and open-weights models, and it is owned by the Swiss nonprofit Proton Foundation. Proton markets the product to both individual users and businesses, positioning it as a privacy-first alternative to mainstream chatbots rather than a narrow consumer tool.

The launch extends a product line Proton first introduced on July 23, 2025, when it unveiled Lumo as an AI assistant built to keep personal data confidential. In August 2025, Proton followed with Lumo 1.1, which it described as smarter and faster while preserving the same encryption model. Lumo 2.0, introduced in Proton’s June 30 blog post, adds a longer list of features that go beyond text generation and into image work and search.

Proton says user testing suggests that for many use cases the qualitative gap between Lumo 2.0 Max and the latest models from OpenAI and Anthropic is no longer perceptible. That is the crucial claim behind the release: if the company is right, privacy could stop being a niche premium and become a buying criterion strong enough to pull users away from the biggest chatbot brands. If it is wrong, Lumo remains a cleanly designed privacy product that still asks users to accept a narrower benchmark than the leaders it is measuring itself against.