Signal Creator Moxie Marlinspike Partners With Meta to Bring End-to-End Encryption to AI Chat

In a move that could reshape the privacy landscape of artificial intelligence, Moxie Marlinspike — the cryptographer behind Signal Protocol and the encryption that protects billions of WhatsApp messages — has announced that his encrypted AI startup Confer will partner with Meta to bring end-to-end encryption to Meta AI.

A Familiar Playbook

The announcement, posted on the Confer blog on March 19, draws a direct parallel to Marlinspike's most impactful work. "Ten years ago, I worked with Meta to integrate the Signal Protocol into WhatsApp for end-to-end encrypted communication," Marlinspike wrote. "That enabled end-to-end encryption by default for billions of people. Now we're going to do the same thing again, for AI chat."

Confer, which launched in January 2026, is an AI chatbot built on open-weight models that claims to offer true end-to-end encryption — meaning not even Confer's own developers can access user conversations. The Verge described it as looking like ChatGPT but working like Signal.

Why This Matters Now

Marlinspike's blog post frames the urgency in stark terms. AI chat applications have become "some of the largest centralized data lakes in history," he writes, containing more sensitive information than any previous technology platform. Users pour their unfiltered thoughts, medical records, financial details, and personal correspondence into AI chatbots — all of which currently flows through unencrypted pipelines.

"Right now, none of that data is private," Marlinspike warns. "It is shared with AI companies, their employees, hackers, subpoenas, and governments."

The partnership means Meta's frontier AI models — among the most capable in the world — would be paired with Confer's privacy technology. Confer will continue to operate independently while Marlinspike works to make its encryption the foundation of Meta AI products beyond basic chat.

The Technical Challenge

Encrypting AI conversations is fundamentally harder than encrypting messaging. With Signal-style messaging, the encrypted data simply passes through servers without being read. But AI models need to process the content of messages to generate responses. Confer has addressed this through what it calls private inference — a system that allows AI models to process encrypted queries without exposing the underlying data.

The approach builds on Confer's earlier work using passkey-based encryption that ties access exclusively to the user's device credentials. No master key exists on any server.

Industry Implications

If successful, the integration could put significant pressure on competitors. OpenAI, Google, and Anthropic all retain varying degrees of access to user conversations, typically for safety monitoring and model improvement. A Meta AI with provable end-to-end encryption would be a first among major AI platforms.

However, the partnership also raises questions. Privacy advocates may wonder whether Meta — a company with a mixed track record on user data — can credibly deliver on encryption promises, even with Confer's technology. Others may question how end-to-end encryption interacts with AI safety monitoring and content moderation.

Key Takeaways

• Moxie Marlinspike (Signal Protocol creator) is integrating Confer's encryption technology into Meta AI
• The partnership mirrors his 2016 work bringing end-to-end encryption to WhatsApp
• Confer uses private inference to process AI queries without exposing user data
• Confer will remain an independent entity while also powering Meta AI's privacy layer
• This could be the first major AI platform to offer true end-to-end encryption by default
• The move puts competitive pressure on OpenAI, Google, and Anthropic on privacy