Proton’s Lumo: A Privacy-First AI Assistant for Security-Conscious Users

Proton, the company behind privacy-focused services like Proton Mail and Proton VPN, has launched Lumo, an AI assistant designed with end-to-end encryption and a strict no-logs policy. Unlike mainstream AI tools, Lumo does not store user prompts or conversations, positioning itself as a solution for those concerned about data privacy in AI interactions¹. The tool is now available on iOS and Android, with a free tier offering limited queries and a paid plan for unlimited access².

Key Security Features

Lumo distinguishes itself with several privacy-centric features. All chats are protected by zero-access encryption, meaning only the user can decrypt saved conversations. Proton cannot access or log chat histories, ensuring compliance with GDPR requirements³. The AI runs on open-source models like Mistral and OLMO 2, hosted on European servers to avoid jurisdiction-related data risks. A unique Ghost Mode allows temporary chats that disappear after closing, while Proton Drive integration enables secure analysis of encrypted files without storing them¹.

Technical Implementation

Lumo’s architecture avoids reliance on third-party APIs, reducing exposure to supply-chain vulnerabilities. Web searches are disabled by default to prevent accidental data leakage, and TLS encryption secures data in transit². The service offers anonymous usage (25 queries/week) or authenticated free accounts (100 queries/week). Paid users ($9.99/month) gain unlimited chats, file uploads, and extended history—all encrypted and deletable at any time³.

Relevance to Security Professionals

For security teams, Lumo provides a safer alternative for tasks like document summarization or code generation, where sensitive data might otherwise be exposed to third-party AI training datasets. Its open-source model stack allows for independent audits, addressing concerns about opaque AI systems¹. Proton’s comparison table highlights Lumo’s advantages over Gemini, Copilot, and OpenAI in data protection, making it a viable option for handling confidential information³.

Conclusion

Lumo represents a significant step toward privacy-respecting AI, though its effectiveness in complex enterprise workflows remains to be tested. Security teams should evaluate its trade-offs between functionality and isolation, particularly for high-risk environments. Proton’s commitment to transparency and encryption sets a benchmark for ethical AI development.