The person who put end-to-end encryption into WhatsApp is now doing the same thing for AI chatbots. Moxie Marlinspike, creator of the Signal Protocol, announced that his startup Confer will integrate its encrypted AI technology into Meta AI.
Confer launched in January 2026 as a privacy-first AI chatbot. It works like ChatGPT or Claude on the surface, but encrypts your prompts on your device before sending them to the server. The AI processes your request inside a secure environment, and the company itself never sees the unscrambled content of your conversations. That means your chats cannot be used for model training or ad targeting, because the host literally cannot access them.
What Confer Brings to Meta AI
In a blog post, Marlinspike wrote that the integration will "combine the most private AI chat technology in the world with the most capable AI models in the world." The privacy layer will serve as a foundation across Meta's AI products, not just a single app.
The details remain thin. There is no rollout timeline, no list of specific Meta products getting the upgrade, and no technical explanation of how end-to-end encryption will work with AI inference (the process of actually generating a response from a model) running on Meta's servers at massive scale. That last question is the hard one. Encrypting a text message between two people is a solved problem. Encrypting a conversation where a cloud-based AI model needs to read your prompt, reason about it, and produce a response is a fundamentally different challenge.
The Track Record Matters
Marlinspike has done this exact playbook before. He built the Signal Protocol, open-sourced it, then licensed it to WhatsApp in 2014. That single integration put strong encryption in front of over two billion people who never had to think about it. If Confer pulls off the same move with Meta AI, the privacy floor for AI conversations rises dramatically without users needing to switch tools or change behavior.
Confer currently runs as a standalone product with a free tier (20 messages per day, five active chats) and a $35/month unlimited plan using open-weight models. How that pricing or product changes after the Meta deal is unclear.
The biggest question is whether "encrypted AI" at Meta's scale will carry the same guarantees as Confer's standalone product, or whether practical compromises will weaken the privacy model. Marlinspike earned trust by shipping Signal with no compromises for over a decade. Whether that standard survives inside Meta's infrastructure is the thing to watch.