- WhatsApp launches an incognito mode for chatting with Meta AI without saving conversations or using them to train models.
- The function is based on Private Processing technology and trusted execution environments so that neither Meta nor third parties can read the messages.
- Chats are temporary, text-only, disappear when you log out, and are intended for sensitive inquiries.
- Meta prepares Side Chat, or parallel chat, to ask Meta AI for help privately within conversations without the other participants seeing it.
WhatsApp has taken another step in its strategy to integrate artificial intelligence into messaging without abandoning its image as a privacy-focused service. The company has introduced a new "incognito mode" for chatting with Meta AI , designed for those who want to take advantage of AI but prefer that these conversations leave no trace.
This new type of chat promises private, temporary, and protected interactions, even from Meta itself . It arrives at a time when more and more users are using chatbots as a kind of digital confessional to discuss health, money, work, or personal relationships, and when regulatory pressure on data use is particularly intense in Europe.
What exactly is incognito mode for talking to Meta AI on WhatsApp?
The so-called Incognito Chat, or incognito chat with Meta AI within WhatsApp , is a special one-on-one conversation mode with Meta's artificial intelligence assistant. Unlike a regular AI chat, here messages are processed in a secure environment and disappear automatically when the session ends.
As WhatsApp has explained, conversations are not stored, used to train models , or linked to the user's history. As soon as the chat is closed, the phone is locked, or the app is exited, the session ends and the AI loses all context, so it cannot be resumed later as if it were a continuous thread.
The idea is to offer a space distinct from the regular chat with Meta AI: something more like a one-off consultation than a permanent history . This way, anyone wishing to ask a sensitive question can do so without that information being saved, at least according to the company's official promise.
This proposal places WhatsApp in the same trend as other generative AI services, such as ChatGPT or Claude, which already offer private or history-free options. The difference is that this proposal attempts to integrate this ephemeral experience directly into a mass messaging app , widely used in Europe and in markets where regulatory compliance is closely monitored.
How does this new AI-powered private chat work?
From the user's point of view, it works relatively simply: just open a chat with Meta AI in WhatsApp —either from the search bar or from the list of conversations— and activate the incognito mode option that will appear in the assistant's window.
Once activated, a private and temporary session is created . Everything written in that chat is processed through WhatsApp's "Private Processing" infrastructure and, by default, is deleted when the user closes the conversation, exits the application, or locks their phone.
This ephemeral behavior means that Meta AI does not maintain long-term contextual memory in this mode. Each session is treated as an isolated block. In other words, it's ideal for making one-off queries, but not for building a history of projects or long-term conversations.
For now, Incognito Mode is limited to text-only interactions . It does not yet support sending images or files, nor does it allow the AI to generate visual content in this private setting. Furthermore, restrictions have been added so the assistant can refuse to respond to harmful or particularly sensitive requests, and a minimum age of 13 is required to use these features.
Meta has indicated that incognito mode won't be limited to WhatsApp: it will also arrive in the standalone Meta AI app in the coming months. However, the company hasn't yet provided a firm date for global rollout or detailed whether there will be regional variations, such as in the European Union, where the regulatory framework for artificial intelligence and data protection is stricter.
Private Processing technology: what it really promises in terms of privacy
The key difference between this incognito mode and other services is the emphasis on the fact that neither Meta nor WhatsApp can read what is written in these chats . To uphold this promise, the company relies on an infrastructure it has been developing for some time: Private Processing.
Private Processing relies on the use of Trusted Execution Environments (TEEs) , a technology that allows data to be processed on servers in an isolated and encrypted manner. The idea is that the message content travels end-to-end encrypted to this secure environment, where the AI models are executed, and that neither administrators, external systems, nor Meta itself can directly inspect the information.
As the company explained, these environments function like a kind of "sealed chamber" within the cloud : data enters to be analyzed by AI, a response is generated, and then the data is deleted. To reinforce this design, Meta claims to use technologies such as Oblivious HTTP and anonymous routing systems that hide the user's IP address, as well as anonymous credentials that verify the request's legitimacy without revealing the user's true identity.
WhatsApp maintains that this architecture is compatible with its end-to-end encryption in personal chats , and that private processing is a way to add advanced AI features without breaking that promise. In fact, the company already used this infrastructure in tools like "Typing Help" and automatic message summarization, and is now simply extending it to the new Incognito Mode with Meta AI.
Furthermore, Meta has announced its intention to open the system to external audits and publish logs of critical components , so that independent researchers can detect suspicious changes in the software that manages these private conversations. This is an attempt to gain credibility in an area—privacy—where the company has a history of controversy.
Why is Meta launching this incognito mode now, and who is it aimed at?
Context helps to understand this trend. In recent years, chatbots have become a go-to resource for increasingly sensitive topics : from financial questions to health issues or personal and work-related conflicts. It's not uncommon for someone to use AI to draft a complex message, seek advice about a breakup, or decipher a medical report.
WhatsApp product managers have admitted that people are starting to share their most private thoughts with AI . This raises concerns about what happens to that data, especially in regions like the EU where the General Data Protection Regulation (GDPR) and the future European AI framework demand greater transparency and control.
There's also a legal factor. In recent months, lawyers and experts have warned that chatbot conversations could be used as evidence in legal proceedings or internal investigations, depending on how and where they are stored. Offering ephemeral sessions with no history and less data retention can be seen as a preemptive response to these concerns.
From a competitive standpoint, the strategy is clear: privacy is becoming as strong a selling point as the quality of the responses . Facing rivals like OpenAI, Google, and Anthropic—also under pressure in Europe to explain how they handle data—Meta aims to position WhatsApp and Meta AI as tools where one can communicate with relative peace of mind, at least in this specific mode.
The target audience for incognito mode is primarily users who already used AI with some reservations or who avoided it altogether for fear of leaving a trace. Although the company emphasizes the system's technical security, it also recommends not sharing extremely sensitive data , such as passwords, physical addresses, or banking information, even in these private chats.
Limits, restrictions, and open questions about incognito mode
Despite its ambitious promise of privacy, the new mode comes with several important limitations and nuances . The first is functional: for now, it only allows text, without sending images or files and without generating visual content from user requests.
The second point relates to the responsible use of AI. WhatsApp indicates that Meta AI may refuse to answer certain queries it deems problematic, dangerous, or requiring professional attention. In matters such as mental health, for example, the assistant may suggest consulting a specialist instead of providing direct instructions.
There's also a less obvious limitation: the lack of a log can complicate later review of interactions . If a conflict arises or you want to verify exactly what was said in an incognito chat, there will be no history to refer to; in such cases, it's helpful to remember guides for removing traces and managing privacy on other services.
Furthermore, although Meta insists it cannot read these conversations, the company remains under intense scrutiny for its record regarding personal data . Past cases, such as the use of content to train models or the access of subcontracted personnel to sensitive information in other products, lead some members of the public to view these types of announcements with caution.
Finally, there are questions about exactly how Incognito Mode will be implemented in Europe . The company mentions a gradual global rollout, but hasn't yet clarified whether the feature will have specific characteristics in the EU or in countries with regulations governing the storage and processing of AI data. It wouldn't be surprising if it arrives with adjustments or additional conditions to comply with EU regulations.
Side Chat or parallel chat: ask the AI for help without anyone else seeing it
Alongside Incognito Mode, Meta has previewed another feature that will also rely on Private Processing: Side Chat . This option allows users to consult Meta AI within a WhatsApp conversation—whether individual or group—without the other participants seeing either the question or the answer.
Until now, if someone wanted to use AI in the middle of a chat, they had to label messages and make the assistant's intervention visible to everyone. This is rather inconspicuous when the goal is, for example, to better understand a message, translate it, ask for help writing a reply, or clarify personal doubts without sharing them with the group.
With Side Chat, the idea is that the user can invoke Meta AI as a kind of background advisor . The assistant would have access to the conversation context (again, through the same private processing infrastructure), but the responses would only appear to the person requesting them.
Meta assures that this parallel chat will also be protected by its privacy system , with messages processed in isolated environments and a promise that neither the company nor third parties can read the content. This aims to balance the real-time utility of AI with a greater degree of discretion within the chats themselves.
Taken together, the new incognito mode for talking to Meta AI on WhatsApp and the upcoming Side Chat show where messaging is headed: integrating intelligent assistants without abandoning the idea of private conversation . We'll have to see how these promises materialize in their actual rollout, especially in Europe, but the direction is clear: the AI race is no longer just about who responds best, but also about who best convinces users that they can do so without reading their entire lives.