Meta introduced Incognito Chat for its Meta AI on WhatsApp and the Meta AI app, enabling users to interact with AI in a secure enclave where Meta cannot read or store the conversations, addressing widespread privacy concerns with AI chatbots.
- Conversations processed in encrypted enclave unreadable by Meta
- Chats deleted by default, no history saved on servers
- Sidechat feature coming to enable AI help within WhatsApp conversations
What happened
Meta has launched Incognito Chat mode for its Meta AI assistant on WhatsApp and the Meta AI app, utilizing a secure processing environment called Private Processing. This system allows AI to interact with user messages inside encrypted Trusted Execution Environments on Meta’s servers, making the content inaccessible to Meta itself.
The conversations in this mode are deleted once the chat session ends, and no server-side record is retained. This approach contrasts with many other AI chatbots that store conversation data by default and can access user interactions. Meta has also released a technical whitepaper on the cryptographic design underlying this new privacy feature.
Why it matters
AI chatbots are increasingly used for sensitive queries, raising significant privacy concerns since many companies retain chat logs or analyze conversation data. Meta’s Incognito Chat aims to eliminate this risk by ensuring AI interactions are processed in a way that even Meta cannot see or store the information.
This innovation helps reconcile WhatsApp’s longstanding commitment to end-to-end encryption with the functionality of AI assistants, addressing tensions between privacy demands and AI capabilities. It also marks a notable step in building user trust by minimizing data exposure risks from AI interaction.
What to watch next
A related feature called Sidechat will soon be added, allowing users to summon AI assistance directly within ongoing WhatsApp conversations. In this mode, the AI can access the chat context without revealing its responses to other participants, extending AI’s utility while maintaining privacy safeguards.
The effectiveness of this enclave-based privacy approach will be closely scrutinized by security researchers and regulators, especially as Trusted Execution Environments have known vulnerabilities. Meta’s invitation for external review and timing amid internal challenges highlights the strategic importance of this privacy-focused AI rollout.