Signal’s creator, privacy-focused cryptographer Moxie Marlinspike, is now applying his experience securing messaging to artificial intelligence by launching an open-source project called Confer, an AI assistant designed to keep user conversations private through end-to-end encryption so even the platform can’t read them — a direct response to concerns about mainstream AI platforms storing and exposing sensitive user data. Various reporting highlights that Confer runs on verifiable open-source software that prevents even operators, hackers, or law enforcement from accessing user prompts and responses, drawing a clear line between this initiative and centralized data collection by dominant AI providers. Marlinspike frames the effort as bringing the kind of privacy protections that have become standard in messaging (through Signal) into the AI space, promoting user autonomy and reducing corporate control over personal information. While major AI companies haven’t signaled a similar push toward end-to-end encrypted AI, Marlinspike’s move underscores a broader debate on data privacy and user rights in next-generation AI tools.
Sources:
https://arstechnica.com/security/2026/01/signal-creator-moxie-marlinspike-wants-to-do-for-ai-what-he-did-for-messaging/
https://www.theverge.com/news/861180/the-creator-of-signal-messenger-is-taking-on-chatgpt-next
https://confer.to/blog/2025/12/confessions-to-a-data-lake/
Key Takeaways
• Moxie Marlinspike has launched Confer, an AI assistant built with end-to-end encryption to protect user interactions from being accessed by providers or third parties.
• Confer is open-source and cryptographically verifiable, positioning it as an alternative to current AI tools that log and retain user data.
• This initiative reflects a growing conservative emphasis on individual privacy rights and skepticism of centralized tech platforms’ handling of personal information.
In-Depth
In the world of artificial intelligence, where data collection and retention have become the industry norm, Moxie Marlinspike’s latest move is a stark, privacy-centric challenge to the status quo. Best known for creating Signal, the gold standard in encrypted messaging, Marlinspike is now attempting to transpose his principles into the AI domain with a project called Confer. The goal is simple in theory but ambitious in practice: build an AI assistant that treats users’ interactions with the same confidentiality afforded to encrypted texts and calls.
Traditional AI platforms, including popular large language models, often require user data to be logged, stored, and analyzed to improve performance. That creates a data “lake” where sensitive information can be subpoenaed, leaked, or otherwise exposed. Marlinspike’s Confer flips this model on its head by ensuring that everything a user types is encrypted on their device and only decrypted in controlled environments that Confer itself cannot surveil. By open-sourcing the entire stack, the project invites public verification of its privacy claims rather than requiring blind trust in a corporation.
This privacy-first approach resonates with a broader conservative perspective on individual rights and skepticism toward centralized digital authorities. It argues that users should retain control of their data and that technology should serve individuals, not harvest them. Whether Confer gains traction against established AI giants remains to be seen, but it’s already shifted the conversation toward a future where secure and private AI is not just an aspiration but a realistic alternative.