Signal’s Creator Is Helping Encrypt Meta AI
The creator of Signal's encryption protocol will help Meta AI protect billions of daily chatbot conversations from snooping.
Moxie Marlinspike, the renowned privacy advocate who created the Signal messaging app and its widely adopted encryption protocol, announced this week that his new AI platform Confer will begin integrating its privacy technology into Meta's AI systems. The collaboration aims to bring end-to-end encryption—the gold standard for secure messaging—to the billions of daily conversations users have with Meta's AI chatbots. Currently, these AI interactions are not encrypted, allowing companies like Meta to access conversations for model training and exposing data to employees, hackers, and government subpoenas.
This partnership builds on Marlinspike's previous work with Meta-owned WhatsApp, where he helped roll out end-to-end encryption to over 1 billion accounts in 2016. The new initiative specifically targets Meta AI, the chatbot integrated into WhatsApp and other Meta platforms, which currently lacks the same privacy protections as individual user chats. While technical details remain scarce, the goal is to create a system where users can access "the full power of AI along with the full privacy of an encrypted conversation," according to Marlinspike's blog post.
Cryptography experts have cautiously praised the initiative, with NYU researcher Mallory Knodel noting it would be "great for people using chatbots that use Meta AI to have confidentiality and privacy within that exchange." Cryptographer JP Aumasson called Confer "probably the best private AI solution, all things considered," while acknowledging the platform lacks documentation about its architecture and threat model. The collaboration represents a significant step toward addressing the fundamental privacy challenge in AI: how to provide powerful generative capabilities without compromising user data security.
- Signal creator Moxie Marlinspike will integrate his Confer platform's encryption into Meta AI systems
- The move aims to bring end-to-end encryption to billions of daily AI chatbot conversations currently exposed
- This follows Marlinspike's 2016 work encrypting WhatsApp and addresses AI training data privacy concerns
Why It Matters
Billions of sensitive AI conversations could gain Signal-level privacy protection, preventing data harvesting for model training.