Lumo AI review: Proton's private assistant, read against its own pages

The Morse teamPublished October 25, 2026Facts checked October 22, 20268 min read

Our Lumo AI review, as of October 22, 2026: Proton's Lumo is a capable private assistant that keeps no logs, doesn't train on your chats, and stores your history so Proton can't read it, but it isn't end-to-end encrypted to the model, and Proton says so itself. Lumo 2.0 added Fast and Thinking modes, image generation, cited web search, memory, and Projects, and Lumo Plus costs $9.99 a month billed yearly.

We read Proton's Lumo page, pricing page, business pricing, Lumo privacy policy (last modified September 17, 2026), privacy support article, Projects and Drive help pages, the Lumo security model post, the Lumo 2.0 and Apertus announcements, and the code repositories Proton links to, all on October 22, 2026. Proton deserves credit on two counts. Its security post explains an AI chat's encryption limits more plainly than any other provider's page we've read, and it runs Lumo's models only on servers it controls. Where our Duck.ai vs Lumo vs Brave Leo comparison covers a fact in depth, we give it a line and link there.

Lumo vs Morse at a glance

As of October 22, 2026. Every Lumo cell comes from the Proton page linked in the section below it.

What's new in Lumo since 2.0?

Our three-way comparison covers the Lumo 2.0 launch of June 30, 2026. Since then, Proton added Apertus 1.5 on September 17, 2026, an open model from EPFL, ETH Zurich, and the Swiss National Supercomputing Centre that you pick "from the model dropdown" (Proton). The Lumo 2.0 post also introduced Custom Lumos for repeated tasks and a context window twice as large, and its speed and quality figures are Proton's own: Lumo 2.0 Lite "scores 127% higher than Lumo 1.4" on the Artificial Analysis Intelligence Index, measured only against earlier Lumo versions (Proton).

Which models does Lumo use?

The app shows Lite, Max, and Apertus 1.5, while Proton's privacy article names the models underneath (Proton). None of the pages we read says which named model sits behind Lite or Max, or how Lumo routes a question between them, and Lumo offers no GPT, Claude, or Gemini.

How much does Lumo cost?

As of October 22, 2026, Lumo has four ways in, from Proton's pricing page and business pricing:

• Free, $0. "Limited Max model usage," "Limited messages & chat history," "Limited image generations," and one Project. The FAQ adds web search and "the ability to upload individual small files."
• Lumo Plus, $9.99 a month billed at $119.88 a year, with $12.99 shown struck through. "More Max model usage," more messages and images, unlimited Projects and Custom Lumos, and "multiple large files." A 30-day money-back guarantee applies.
• Lumo Professional, $11.99 per user a month billed at $143.88 a year, adding admin tools.
• Workspace Premium, $19.99 per user a month billed at $239.88 a year, which bundles Lumo with Proton Mail, Calendar, Drive, VPN, Pass, and 3 TB of storage per user.

Both business plans list a 14-day free trial, and $14.99 and $24.99 struck through as their full prices. If you drop from Plus to Free, Proton says your saved chats stay and the Plus features end with the billing period (Proton).

Proton doesn't publish the Free plan's limits as numbers, on the pricing page or in its FAQ, which says only that paid plans have "much higher" limits. A Project holds up to 200 files, and Lumo "usually selects fewer than 5" of them for any one answer (Proton). Free accounts get one Project at a time, and Plus and Professional get unlimited ones. You can link a Proton Drive folder to a Project, and "whenever new assets are added to the linked Proton Drive folder, Lumo will automatically refer to them"; long documents are split into chunks so they can still be indexed. Proton says Lumo handles documents, spreadsheets, code, and PDFs, and "deletes them immediately after processing a response" (Proton).

Is Lumo end-to-end encrypted?

No, by Proton's own account: it says it must "provide the LLM with cleartext messages" and calls the setup "user-to-Lumo (U2L) encryption" (Proton). We cover that model, and the missing enclave check, in our Duck.ai vs Lumo vs Brave Leo comparison. Proton's Projects help page does say "Projects are end-to-end encrypted and synced across your devices" (Proton), but that describes stored Projects. The files you ask about still reach the model in the clear.

What does Lumo's privacy policy add?

Proton's no-logs, no-training, and zero-access history claims are in our comparison of Lumo with Duck.ai and Brave Leo. The Lumo privacy policy, last modified September 17, 2026, adds three things the support article leaves out (Proton):

• Web search reaches outside Proton. With web search on, "Lumo may transmit a simplified version of your request to selected partner APIs," though "the full content of your query is not shared." The partners aren't named, and you can switch search off.
• Abuse checks use your account data. Proton "may use the data mentioned above to detect abusive and fraudulent use."
• Feedback leaves Proton. Apertus feedback "may also be shared with ETH Zurich and EPFL."

Is Lumo open source?

Partly. Proton's Lumo page says "Lumo's code is fully open source, so anyone can verify that it's private and secure," and its "Explore Lumo's code" link goes to Proton's open-source page (Proton). As of October 22, 2026, that page links three repositories, all of them apps: the Android app and iOS app in the ProtonLumo GitHub organization, both GPL-3.0, and the web app inside Proton's WebClients repository (Proton). We found no public code for the model servers, the system prompt, or the routing between models.

That leaves the client verifiable and the server taken on trust. The European Open Source AI Index made the same point in August 2025, calling Lumo "the least open 'open' AI assistant" because "the only open source code we have found is for the Lumo mobile and web apps." Proton answered that "Lumo isn't a model; it uses open models" (European Open Source AI Index).

Where is Lumo based?

In Switzerland. The privacy policy says Lumo is run by Proton AG in Plan-les-Ouates and "governed by the laws and regulations of Switzerland," and that Proton discloses "the limited user data we possess" only on "a binding request coming from the competent authorities" (Proton). Proton's Lumo page says it is "outside of US jurisdiction" (Proton). The pages don't say which countries Lumo's model servers are in, and the policy lists payment and support processors in the United States, among other places.

How Morse handles the step Lumo asks you to trust

Lumo's privacy rests on one server. Proton's model server reads your message in the clear, and nothing outside Proton checks what that server runs. Morse moves that check onto your own device.

• E2EE, for anything you'd otherwise have to trust a server with. Your browser checks the enclave's cryptographic proof itself, against versions pinned in the app, before anything leaves your device, then seals the message so only that enclave can read it. Our servers relay text they can't read, and the reply is decrypted only in your browser. Every E2EE reply has a lock you can open to see each check that passed. E2EE runs open-weight models such as GLM, Kimi, and DeepSeek, and comes with Pro at $19 a month. It rests on the enclave hardware and attestation working as designed, and it skips web search and image generation, so start a chat in another mode when you need to search.
• Anonymous, for the GPT, Claude, and Gemini that Lumo doesn't offer. Our proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider sees the question without knowing it's you. It may keep it for a limited time to check for abuse, under its own policy.
• Private, for open models with nothing kept. Morse has zero data retention agreements with its inference, search, and other providers. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.

Your history is stored only in your browser, encrypted, and Morse's servers don't keep or log your messages. Files you attach go with the message and are never stored by Morse. We don't train AI on your content, sell your data, or show ads (privacy policy). Morse can't recover a chat you clear, so save a copy of anything you need. The same account holds Characters and Studio, with unfiltered models on paid plans.

Choose Lumo or Morse

• Choose Lumo if you already use Proton Mail and Drive, want Projects that read a linked Drive folder, want models run only on servers Proton controls under Swiss law, or want a free assistant you can reach as a guest over Tor.
• Choose Morse if you want your own browser to verify the server before a word leaves it, want GPT, Claude, and Gemini in the same account with your identity stripped, and want history that lives only on your device.

Try a chat from the home page without an account, or start free and move to Pro when you want E2EE.

Morse publishes this comparison. We've tried to keep it fair; if something's wrong or out of date, email hello@trymorse.ai.

Proton and Lumo are trademarks of Proton AG. Morse is not affiliated with or endorsed by Proton AG.