Lumo
AI assistant from Geneva-based Proton that runs on open-source models, stores chats encrypted and does not use your inputs for training.
What it does
Lumo answers questions, summarises texts and helps with writing, with optional web search for current information. You can upload files or link files and folders from Proton Drive, for example to analyse reports or look through code. Projects group chats on one topic, and “Custom Lumos” store recurring instructions. It can also generate and edit images. Under the hood are open-source models such as Qwen and GLM, which Proton runs on its own servers.
Who it suits
Teams that want an all-round assistant but would rather not have confidential content processed by large US providers – especially if they already use Proton Mail or Proton Drive.
Data protection for businesses
Proton is based in Geneva, describes Lumo as built and based in Europe and uses only servers it controls; the vendor does not name a specific data centre. Chats are not used for training, so there is nothing to switch off. Only feedback you choose to give on answers from the Apertus model is shared, anonymised, with Proton and the universities involved. Saved chat history is encrypted so that only you can read it. Proton has a data processing agreement as part of its terms, but it does not name Lumo explicitly. For companies there is Lumo Professional with admin tools, and Workspace Premium with the full Proton suite.
Limits
Even Plus and Professional cap usage of the strongest model, messages and images. The free version allows only one project and small files. There are no desktop apps. Answers can be wrong and need checking.