Moxie Marlinspike, Signal co-founder, launched Confer in December, offering a privacy-conscious AI service that prevents data collection and ad targeting.
Confer resembles ChatGPT or Claude in appearance and function. Its backend design avoids data collection, leveraging open-source methods. Conversations within Confer cannot be used for model training or ad targeting as the host does not access them.
Marlinspike states these protections address the intimate nature of AI services. He observes, “Chat interfaces like ChatGPT know more about people than any other technology before. When you combine that with advertising, it’s like someone paying your therapist to convince you to buy something.”
Privacy in Confer relies on several integrated systems. Messages are encrypted using the WebAuthn passkey system. This standard performs optimally on mobile devices or Macs running Sequoia, though it can function on Windows or Linux with a password manager. On the server side, all inference processing occurs in a Trusted Execution Environment (TEE). Remote attestation systems verify the TEE’s integrity. Within this environment, open-weight foundation models process user queries.
This setup is more complex than standard inference configurations but delivers on Confer’s promise of user privacy. Users can engage in sensitive conversations with the model without information leakage, provided these protections remain active.
Confer’s free tier provides 20 messages daily and five active chats. A premium subscription, priced at $35 per month, offers unlimited access, advanced models, and personalization features. This price exceeds ChatGPT’s Plus plan.
