The person who brought end-to-end encryption to billions of WhatsApp users is now aiming to do the same for AI. Moxie Marlinspike, the cryptographer and founder of Signal, announced this week that his encrypted AI startup Confer will work to integrate its privacy technology into Meta AI — a move that could redefine the baseline for data protection in AI chat.
A Second Act in Messaging Privacy
In 2016, Marlinspike worked with Meta to implement the Signal Protocol in WhatsApp, enabling end-to-end encryption by default for more than a billion users at the time. The move became a landmark moment in mainstream consumer privacy.
Now, Marlinspike says he sees an analogous opportunity — and urgency — in AI. "AI chat apps have become some of the largest centralized data lakes in history, containing more sensitive data than anything ever before," he wrote in a post announcing the collaboration. "We are using LLMs for the kind of unfiltered thinking that we might do in a private journal — except this journal is an API endpoint to a data pipeline specifically designed for extracting meaning and context."
His concern is pointed: insecurities, incomplete thoughts, medical records, financial details, and personal correspondence are all flowing into AI systems — and none of it is currently private. "It is shared with AI companies, their employees, hackers, subpoenas, and governments," he wrote.
How Confer Works
Confer was built on top of open-weight models and uses a passkey-based encryption architecture that ensures not even Confer itself can access user conversations. Marlinspike describes it as being built so that "nobody has access to your conversations but you" — a privacy guarantee that has historically only existed in secure messaging, not AI.
The limitation, Marlinspike acknowledges, is that open-weight models can't match the frontier capabilities of proprietary systems from OpenAI, Google, or Anthropic. That's part of why the Meta collaboration is significant: it pairs Confer's privacy technology with the most capable frontier models Meta is developing.
Confer Remains Independent
Marlinspike was clear that Confer will continue operating as an independent entity. The arrangement is a technology integration, not an acquisition — at least for now. His goal, as stated, is to ensure that as Meta builds more AI products beyond basic chat, "the privacy technology from Confer will be a part of the foundation of everything that is to come."
Why This Matters
The AI industry has largely operated on a model where user data flows freely to providers in exchange for service quality. Privacy-preserving alternatives have existed but remained niche. If Marlinspike can bring encrypted AI to Meta's scale — which spans billions of users across WhatsApp, Instagram, Messenger, and the Meta AI assistant — it would represent a fundamental shift in what users can expect from AI services.
The timing is notable. Meta recently launched its AI support assistant on Facebook and Instagram, and is deploying AI systems across content moderation, search, and social features. Baking privacy guarantees into that infrastructure from the start would be a different approach than retrofitting encryption after the fact — the path most platforms have historically taken.
Whether Confer's technology can scale to meet Meta's infrastructure demands, and whether Meta is genuinely committed to privacy-preserving AI or is simply pursuing a favorable narrative, are questions the industry will be watching closely as this partnership develops.
