'Lumo 1.1' Is Smarter, Faster, And Private, But Less Adaptive Without Surveillance Capitalism

When some consume data like nothing before them did, some others try a more subtle approach.

Ever since ChatGPT burst onto the scene, the world has been locked in what many now call the LLM wars. Tech giants like OpenAI, Google, Anthropic, Microsoft, and Meta have been racing to release bigger, faster, and flashier AI assistants. Each new launch promised more intelligence and more capabilities.

But the trade-off was always the same: users had to surrender their privacy.

Conversations were logged, data was harvested, and the very interactions meant to help people were quietly repurposed for training, profiling, and in some cases, monetization.

It didn’t take long before a countercurrent emerged.

Brave, for example, introduced Leo, its privacy-conscious chatbot designed to process requests without storing conversations. Then in July 2025, Proton, the Swiss company best known for its encrypted email service, entered the arena with a bold declaration: AI doesn’t have to consume data from users to be useful.

It introduced Lumo, an assistant that arrived with a different philosophy: private by design, transparent in its workings, and open source at its core.

And this time, 'Lumo 1.1' has been released.

The update isn’t just a minor refinement but a significant leap forward in performance while preserving the privacy-first promise that defines Lumo.

In a blog post, Proton said that tests show that the version has a 170% improvement in context understanding, making responses sharper and less generic, a 40% improvement in coding capabilities, and a 200% boost in reasoning, allowing Lumo to handle multi-step planning and complex workflows more reliably. Response times are also faster, accuracy is higher, and all of this happens without compromising confidentiality.

Proton also went further in transparency by open-sourcing its mobile apps alongside the web client and publishing the Lumo security model, offering a rare look into how its zero-access encryption works.

Being a chatbot draped without much marketing language, it tries to differentiate itself by providing privacy first.

In other words, Lumo is AI that truly works for the user, not for advertisers or data brokers.

Lumo 1.1 fits neatly into this space by combining trust with the raw capability of a modern AI system.

It’s a reminder that the future of AI doesn’t have to be built on surveillance capitalism. In an industry where speed and performance often come at the hidden cost of control and transparency, Proton is making a different bet—that users will choose power and privacy together when given the option. In doing so, Lumo is carving out its place in the LLM wars not by being louder, but by being smarter, safer, and fundamentally more trustworthy.

Then comes the disadvantages.

Privacy is Proton’s greatest strength, but in the ongoing LLM war, this privacy greatness reveals fundamental weaknesses.

Unlike OpenAI, Anthropic, or Google, Lumo refuses to harvest user data. Whereas tech giants improve their models in part because they retrain constantly on billions of real conversations, turning every interaction into a free lesson. Lumo, by design, cannot do this. By not consuming user data, it limits itself to curated sources and publicly available datasets.

That makes it safer, but it also leaves it at risk of falling behind in adaptability and conversational polish.

The same restraint applies to personalization. ChatGPT, Claude, and Gemini build profiles of their users over time, shaping responses to habits, quirks, and preferences. Lumo, with its zero-access encryption and ephemeral design, cannot keep that kind of memory without compromising its promise of confidentiality.

As a result, its conversations are private but less tailored.

Scale is another challenge.

Running a privacy-first AI with end-to-end encryption, European hosting under GDPR, and open-source verification isn’t cheap. These choices add friction to development and force Proton to move carefully. While competitors roll out flashy features like AI voice companions or real-time multimodal interaction, Lumo evolves more deliberately, balancing progress with strict privacy guarantees.

Still, these limitations are not flaws. They are the inevitable byproducts of its philosophy.

By choosing privacy over surveillance capitalism, Lumo may sacrifice some adaptiveness, but it can shine in areas where others can never compete.