Proton’s new privacy-first AI assistant encrypts all chats, keeps no logs

Proton, known for its commitment to privacy-centric productivity solutions, has unveiled Lumo, a new AI assistant designed with a strong emphasis on user data protection. This innovative chatbot ensures that no conversation logs are retained, employing end-to-end encryption for chat storage and offering a unique ghost mode that allows discussions to vanish upon closing the window. Accessible through a web client and available on both Android and iOS platforms, Lumo does not require users to create an account to interact with the assistant. Users can even upload files for the chatbot to analyze and respond to, and those with a Proton Drive account can seamlessly connect it to access their stored cloud files. While Lumo is equipped with web browsing capabilities, it may not always deliver the most up-to-date search results. Proton is keen to reinforce its privacy-first approach, noting that Lumo is built on open-source models. The company commits to using these models solely for research and development purposes, ensuring that user data is never utilized for training. Lumo operates using zero-access encryption—a method prevalent across Proton's products—allowing users to store their chat histories in a manner that can only be decrypted on their own devices. In its announcement, Proton highlighted its European headquarters, asserting that this location provides it an advantage over AI firms based in the U.S. and China when it comes to privacy issues. "Lumo is powered by open-source language models and functions from Proton's European data centers, offering significantly greater transparency than other major AI assistants," the company stated. Unlike other AI services that collaborate with firms like OpenAI, Proton emphasizes that Lumo does not share user queries with third parties. This launch adds to Proton's growing portfolio of AI tools, building on its previous introduction of an AI writing assistant for its Mail service, which also operates directly on user devices.