We are the Frontier
U.S.

Moxie Marlinspike Launched Privacy-Focused Chatbot Confer

The service employs server-side enclaves to ensure user conversations remain inaccessible to operators.

Updated on Sept. 22, 2026 in Artificial Intelligence

Bold flat-color editorial illustration of a stylized computer processor, using navy, cream, and orange to represent secure AI infrastructure.
Moxie Marlinspike launched Confer, an AI chatbot service that secures user conversations using Trusted Execution Environments to block operator access. AI Illustration. Upload story photo >

Live Poll

Do you trust AI platforms to keep your personal conversations private?

In December 2025, Moxie Marlinspike released Confer, an AI chatbot designed to provide assistance while structurally preventing operator access to user data. The service operates via an open-source technical framework that is now available to the public.

Why it matters

The platform was built to enable AI interaction without compromising data privacy, creating a model for secure inference. This approach addresses growing user concerns regarding the transparency and security of large language model training and processing.

Confer restricts free users to 20 messages and 5 active chats per day. Premium access requires a $35 monthly subscription.

The players

Moxie Marlinspike

A security researcher and software engineer known for his prior work on the Signal messaging protocol.

Meta AI

The artificial intelligence research and product division of Meta Platforms, which develops large language models for social and professional applications.

The details

Confer secures data by utilizing WebAuthn passkeys—a cryptographic standard for passwordless authentication—to encrypt conversations directly on the user device. The service performs inference inside Trusted Execution Environments, which are secure, isolated areas of a processor designed to protect data during execution. This architecture ensures that even the system operator cannot view the contents of a conversation, regardless of the AI model being used.

Timeline

  1. • December 2025: Moxie Marlinspike launched the Confer chatbot.

  2. • March 2026: Meta AI is expected to integrate Confer's privacy architecture.

The Tech Race

Confer applies existing hardware-enclave research to the generative AI stack, marking a departure from cloud-hosted models that require decrypted data in memory. This development follows a pattern established by the secure computing community to move sensitive processing away from central server access.

Users can currently access the service via the web, provided they have hardware capable of supporting WebAuthn protocols. The primary workflow change for users is the necessity of managing passkeys to maintain encrypted sessions, which differs from traditional email-and-password logins.

The takeaway

Confer signals a shift toward privacy-first AI where architecture governs security rather than relying solely on company policy. Readers should monitor the March 2026 Meta AI integration to see how this hardware-enclave approach scales across massive consumer platforms.

What happens next

Meta AI is scheduled to integrate the Confer privacy architecture into its own systems in March 2026.

Further reading

For more on the development of secure model processing, visit Artificial Intelligence.

More information

To explore the platform's technical framework and service tiers, visit the Confer AI service website.

Source note: This article includes information reported by Crypto Briefing.

Live Poll

Do you trust AI platforms to keep your personal conversations private?