Signal creator Moxie Marlinspike wants to do for AI what he did for messaging

Confer, the open source AI assistant unveiled by Signal creator Moxie Marlinspike, aims to do for AI what Signal did for messaging by making user data unreadable to the platform operator, hackers, law enforcement, or any other party other than account holders.

The service runs entirely on open source software that users can cryptographically verify is in place, and data and conversations originating from users and the resulting responses from the LLMs are encrypted in a trusted execution environment, which prevents even server administrators from peeking at them. Conversations are stored in the same encrypted form, with the key remaining securely on users’ devices, and Confer uses passkeys to enable two‑factor authentication and device‑bound login.

The private key enables encryption of input and output, allows syncing across devices without compromising privacy, and supports forward secrecy to ensure past chats remain unread even if a key is compromised. According to Dan Goodin, Confer’s architecture includes a trusted execution environment on servers, remote attestation, and transparency logs for releases, with native support on macOS, iOS, and Android, while Windows and Linux have limited or no native support.

Emphasising the data‑privacy challenge of AI models, the article notes that AI assistants are “archnemesis” data privacy because they rely on massive data collection from many sources.