Proton’s Lumo 2.0 adds image tools, faster search and memory

Proton launched Lumo 2.0 on June 30, 2026, adding image recognition, image generation and image editing to its privacy-first chatbot as it tries to turn privacy from a promise into a competitive advantage. The update also brought a new thinking mode for more complex questions, live web search with source citations and expanded Projects with user-controlled persistent memory.

The company said the revised assistant responds to most queries up to 76% faster than the previous version. Proton also said Lumo 2.0 Lite scored 127% higher than Lumo 1.4 on the Artificial Analysis Intelligence Index, while Lumo 2.0 Max scored 240% higher. Those gains matter because consumer AI has increasingly become a race on speed, quality and multimodal features, not just on privacy branding.

Andy Yen, Proton’s founder and chief executive, said the system was “re-engineered from the ground up” and argued that users no longer need to choose between powerful AI capabilities and meaningful privacy protections. That claim is central to Lumo’s positioning. Proton says conversations are protected with zero-access encryption, meaning even Proton cannot read stored chats, and it says the product does not use data to train AI models.

Lumo first launched on July 23, 2025 as a free assistant that required no account to start chatting. Proton framed it then as part of a broader privacy ecosystem built in Europe, supported by community funding rather than advertising, owned by the nonprofit Proton Foundation and compliant with GDPR. The company says that ecosystem is trusted by more than 100 million people worldwide.

The question now is whether the new tools materially change the trade-off facing users. Image generation, editing and persistent memory bring Lumo closer to mainstream assistants from OpenAI and Anthropic, while the privacy architecture remains Proton’s main point of differentiation. Yet the practical edge depends on how much of the experience stays genuinely private, how useful the encrypted memory proves in day-to-day use and whether users see enough performance to make privacy a deciding factor rather than a niche preference.