WhatsApp launches Incognito Chat with Meta AI and promises full chat privacy

WhatsApp is launching Incognito Chat mode for communication with Meta AI, which the company calls completely private. According to Meta, even the platform itself cannot read requests and responses: they are processed in a protected environment and disappear after the session ends.

How It Works

The new mode is built on top of Private Processing technology, which WhatsApp introduced a year ago for AI features like message summarization and writing suggestions. The idea is to give users access to large models without breaking the messenger's traditional promise of conversation confidentiality. Instead of regular server processing, requests go to an isolated execution environment where they cannot be viewed even from within Meta's infrastructure.

Essentially, Meta is trying to reproduce the properties of an end-to-end approach for tasks that cannot be entirely performed on a smartphone. WhatsApp head Will Cathcart compared this to a giant phone for AI that doesn't have a password even for the company itself. Meta claims that in such a scheme, it only sees the fact of using the feature, but not the content of the conversation.

Incognito Chat began rolling out on May 13, 2026 in WhatsApp and the Meta AI app, with global rollout taking several months.

What Users Get

From an interface perspective, this is a separate temporary session with Meta AI, designed for questions that people typically don't want to leave in regular chat history: from financial topics to career advice and personal matters. Meta is betting that a private mode will lower the main barrier to using AI in a messenger.

• Conversation is visible only to the user
• Messages are not saved by default and disappear after exiting the session
• Currently the mode works only with text
• For answers on current topics, AI can anonymously access web search; this option can be disabled
• In coming months, Meta promises Side Chat — private AI tips on the current conversation without interfering with the main correspondence

Separately, Meta says it will later add image processing and voice recognition. If this appears in the same Private Processing loop, the company will be able to expand scenarios from simple questions to analyzing photos, voice messages, and group discussions. For WhatsApp, this is an important test: users will either accept AI as a built-in tool, or finally decide that an assistant has no place in personal correspondence.

What Questions Remain

Meta promises external audits, vulnerability reporting, and code review by independent specialists. This matters because the entire structure relies not only on cryptography, but also on trust in how exactly the company deploys and maintains the system in the cloud. Johns Hopkins cryptographer Matt Green, who is familiar with the Private Processing architecture, believes the idea is generally workable.

"If you want to talk to AI in a way that no one, including

Meta, can see the conversation, this scheme should work."

But even with such an architecture, risk doesn't disappear. Any protected cloud system remains an attractive target for attacks, and the cost of an error here is especially high: users can share medical, financial, and family details with the bot. Skepticism is also added by the general context around Meta. That same week, the company completely removed opt-in end-to-end encryption from Instagram Direct, so its approach to privacy doesn't look equally strict across all products.

What This Means

If Incognito Chat works as Meta promises, it could become the first truly mass example of a private AI chat within a messenger. For the market, this is a signal that the next competition will go not only for model quality, but for who can more convincingly prove: your requests to AI will not remain on someone else's server in readable form.