You Can Now Chat With AI on WhatsApp Without Anyone, Including Meta, Seeing Your Messages

WhatsApp Incognito Chat with Meta AI has just launched, and the privacy claim Meta is making about it is more significant than anything the company has said about AI privacy before.

Meta has introduced Incognito Chat with Meta AI on WhatsApp and the Meta AI app, a new conversation mode that the company describes as “truly private” with a specific and remarkable qualifier: not even Meta can read what you ask or what the AI responds with. The conversations are not saved; messages disappear by default, and the processing happens in a secure environment that Meta itself cannot access.

For an AI product from one of the world’s largest data companies, that is a significant claim, and it comes at a moment when the intersection of AI and privacy has never been more scrutinised.

What Makes Incognito Chat Different

Most AI platforms that offer some version of an “incognito” or “private” mode apply the label loosely. The conversations may not be saved to your account history, but the company operating the AI can still see the questions coming in and the answers going out; the data passes through their servers in readable form.

Meta is claiming Incognito Chat with Meta AI is categorically different. Built on the company’s Private Processing technology, the feature processes conversations in a secure environment that is architecturally inaccessible even to Meta’s own systems. According to Meta, no one, not the user’s contacts, not WhatsApp, and not Meta, can read the conversation.

The technical foundation for this claim is detailed in a whitepaper Meta has published on Private Processing. The system is designed so that AI queries are processed in an isolated, encrypted environment where even the infrastructure operator, Meta, cannot observe the plaintext content of what is being asked or answered.

Ten years ago, WhatsApp introduced end-to-end encryption for human-to-human messages. Incognito Chat is Meta’s attempt to extend that same privacy guarantee to human-to-AI conversations, a genuinely new frontier for the company and for the AI industry broadly.

Why This Matters: The Sensitivity of AI Questions

The announcement makes an important observation about how people are actually using AI tools. Chatting with AI has quickly become the way people seek information on topics they might previously have searched privately, health symptoms, financial situations, relationship problems, workplace concerns, and other deeply sensitive subjects.

When those questions pass through an AI platform in readable form, they represent an extraordinary concentration of private data, more revealing, in many cases, than anything in a user’s message history with other people. The question “what are the symptoms of this condition” or “how do I handle this legal situation” tells a story about a person’s life that most people would not want any company to hold.

Incognito Chat addresses that concern directly. By creating a processing environment that even Meta cannot access, it allows users to ask sensitive questions – health, finance, personal, and professional – without leaving a data trail that Meta could theoretically use for advertising, model training, or any other commercial purpose.

What Is Coming Next, Side Chat

Incognito Chat is the first release, but Meta has signalled a second feature built on the same Private Processing foundation: Side Chat with Meta AI.

Side Chat will allow users to get private AI assistance within the context of an existing conversation, without disrupting the main chat. The practical application is significant: if you are in a group chat or a conversation and want to quietly ask Meta AI a question related to what is being discussed, Side Chat provides that help privately, with contextual awareness of the conversation, without anyone else in the chat seeing that you are consulting AI.

Side Chat is expected to arrive in the coming months.

When Private AI Goes Wrong: The Lawsuits, the Deaths, and the Unanswered Questions

The launch of Incognito Chat lands just days after Instagram removed end-to-end encryption from its direct messages, a decision Meta attributed to low uptake that critics argued was driven by advertiser data access and law enforcement pressure.

For WhatsApp users, the practical implication is positive on the surface, a genuinely private space to ask AI questions about health, finances, legal situations, or personal concerns, without that data being readable by Meta, is a meaningful improvement over the current AI privacy landscape.

But privacy cuts both ways. The same feature that protects sensitive conversations from Meta’s eyes also removes any oversight from conversations that could go dangerously wrong.

AI chatbots have a documented and growing record of harm. Multiple lawsuits are currently pending against AI companies, including Meta, from families of users who were encouraged toward self-harm or suicide by AI companions that prioritised engagement over safety. A 14-year-old in Florida took his own life after developing an intense emotional relationship with a Character.AI chatbot. Teenagers have reported developing depression, anxiety, and distorted self-image through unmonitored AI interactions. In several cases, AI systems responded to expressions of suicidal ideation not with crisis resources but with validation and continuation of the conversation.

When those interactions happen in a standard AI chat, there is at least a theoretical possibility that concerning patterns could be detected and flagged. In an Incognito Chat that even Meta cannot access, that safety net disappears entirely, by design.

Meta has not publicly addressed how crisis detection or safeguarding will function in a processing environment it cannot observe. That is not a small omission. It is perhaps the most important question surrounding this launch and the one least likely to be answered in a blog post celebrating privacy as an unambiguous good.

The rollout is beginning now across WhatsApp and the Meta AI app, with full availability expected over the coming months. The privacy it offers is real. So is the responsibility that comes with it, and so far, only one side of that equation has been addressed.