WhatsApp is rolling out something that sounds almost too good to be true in the age of AI-powered everything: a way to chat with Meta AI where nobody, not even Meta itself, can see what you’re asking.

Announced on May 13, 2026, Incognito Chat with Meta AI is WhatsApp’s latest attempt to bring artificial intelligence into messaging while keeping privacy intact. And this isn’t just another incognito mode in name only. Unlike other apps that slap a private label on features while still collecting your questions and responses behind the scenes, WhatsApp says this is genuinely invisible to everyone, including the company that built it.

The timing makes sense. AI chatbots have become the go-to for all kinds of questions, from the mundane to the deeply personal. People ask AI about financial situations, medical symptoms, work problems, relationship advice, and things they’d never post publicly or share with another person. The problem is that most AI services see everything you type, store it, and sometimes use it to train future models. WhatsApp is betting that people want AI help without that trade-off.

Incognito Chat runs on what WhatsApp calls Private Processing, a technology the company introduced earlier this year. When you start one of these sessions, your conversation happens in what Meta describes as a secure, isolated environment. Your messages are processed there, but they’re never saved. By default, everything disappears once you close the chat. The session also ends if you lock your phone or close the app, and Meta AI loses all context from that conversation.

The feature uses Oblivious HTTP, a protocol that obscures your IP address during the exchange, making it harder to link requests back to your identity or location. Meta says it has no access to the content of these chats, period. That’s a significant claim, especially for a company that’s faced intense scrutiny over data collection practices in the past.

For now, Incognito Chat is limited to text only. You can’t upload images or files in this mode. Meta’s AI also includes safety guardrails, meaning it will decline inappropriate requests or steer conversations in a different direction when needed. The feature is rolling out gradually on WhatsApp and the standalone Meta AI app over the coming months.

WhatsApp has been walking a fine line with AI. The app’s entire reputation rests on end-to-end encryption, a feature it brought to billions of users a decade ago. Adding AI tools that require cloud processing threatened to undermine that trust. Private Processing is Meta’s answer to that tension, a way to offer AI capabilities without breaking the encryption promise that defines WhatsApp.

This isn’t the only privacy-focused AI feature in the pipeline. WhatsApp also teased Side Chat, another Private Processing-powered tool launching in the coming months. Side Chat will let you privately ask Meta AI questions about an ongoing conversation without other participants seeing the interaction. Right now, if you want to ask Meta AI something about a group chat, you have to tag a message, and everyone can see your question and the AI’s response. Side Chat solves that by letting you pull Meta AI aside for help without disrupting or exposing anything to the rest of the group.

It’s worth noting that regular chats with Meta AI on WhatsApp, the ones that aren’t in Incognito mode, are still not end-to-end encrypted. That’s been a sticking point for privacy advocates. Earlier in 2025, WhatsApp introduced Advanced Chat Privacy, a separate feature that lets you disable certain AI features in chats and prevent conversations from being exported. Advanced Chat Privacy is a per-chat setting that also blocks auto-downloading media and stops anyone in the chat from exporting messages. It’s optional and off by default, but it gives users more control over how AI interacts with their conversations.

Incognito Chat with Meta AI represents a bigger bet. It’s WhatsApp saying that AI doesn’t have to come at the expense of privacy, that you can ask sensitive questions and get help without worrying about who’s watching or what’s being stored. Whether users trust that promise remains to be seen, but at least on paper, this is one of the more aggressive privacy moves in the AI space so far. Meta is essentially building AI features with intentional blindness, creating systems where even the company itself can’t see what’s happening.

The rollout is happening now, so expect to see Incognito Chat appearing in your WhatsApp over the next few months. If Meta follows through on the technical claims, it could set a new standard for how AI assistants handle privacy. If nothing else, it’s a clear signal that the next battle in AI won’t just be about who builds the smartest assistant, but who builds the one people actually feel safe talking to.