WhatsApp's new AI mode erases your chats forever: but experts warn of a hidden danger

WhatsApp has launched a new 'Incognito Chat' feature for its Meta AI, promising that your conversations will never be seen or stored by anyone, including Meta. While the move is a huge step for user privacy, cybersecurity experts have raised concerns about a lack of accountability if the AI gives harmful advice.

What is 'incognito chat'?

WhatsApp is boosting its artificial intelligence features with a new mode designed for total privacy. The feature, called "Incognito Chat with Meta AI," allows users to have temporary, private conversations with the chatbot that disappear once the session is closed. Parent company Meta has assured users that the content of these chats is not visible to the company, is not saved on its servers, and will not be used to train future AI models.

The session ends entirely if the user closes the app or locks their phone, leaving the AI with no memory of the discussion. In May 2025, Meta boss Mark Zuckerberg described it as the "first major AI product where there is no log of your conversations stored on servers". The technology leverages what Meta calls "Private Processing technology," which uses secure digital environments to handle the chats. Will Cathcart, the head of WhatsApp, added that while the tech is not the same as the end-to-end encryption used for user messages, it is "the equivalent" in terms of privacy protection.

A direct response to user fears

The launch of the new feature comes after Meta faced criticism from some users when its AI was first added to WhatsApp last year, with many angry at not being able to turn it off. The company says the incognito mode is a direct response to a "category-wide privacy concern" as more people use AI for sensitive queries about health or personal finance.

"We've heard from a lot of people that they feel some discomfort about sharing [personal] information with the company, yet they want the answers," Cathcart said of the announcement. Currently, most AI companies, including the makers of ChatGPT and Google's Gemini, store some data from chatbot interactions. Outside of premium business accounts, this data is often used to train and improve their AI models. By making these chats completely invisible even to itself, Meta is positioning its AI as a more privacy-conscious alternative.

Experts warn of a 'lack of accountability'

However, the promise of total privacy has sparked new concerns about how the system could hide AI malfunction or abuse. Professor Alan Woodward, a cybersecurity expert at Surrey University, said that while there was a low risk to WhatsApp's existing security, the disappearing nature of the chats could lead to a lack of accountability for the AI’s responses.

The concern is that if a user is given harmful, dangerous, or illegal advice by the AI, the ephemeral messages would mean it would be impossible to find evidence of what was said. This is particularly worrying as many AI companies, including OpenAI and Google, have already been the subject of wrongful death lawsuits. "Personally I think what you ask an AI should remain private as some people ask it very personal matters - but you are placing a great deal of trust in the AI not to lead users astray," he told the BBC. Without a chat history that can be retrieved by the user or Meta, it could be impossible to prove whether an AI interaction led to harm.

Meta's safety 'guardrails'

WhatsApp has said it is aware of these risks. Cathcart stated that the incognito mode would have "guardrails" that would err on the side of caution, meaning the AI would be programmed to refuse requests that could be interpreted as harmful or illegal. To further limit the risk, the feature will initially only process text, with image uploads not supported. The platform has also blocked other AI chatbots from being accessible through its systems, ensuring that the only AI its billions of users can interact with is Meta's own.

A £107 billion gamble on AI

The push for a more private AI is part of a much larger strategy for Meta. According to Susannah Streeter from investment platform Wealth Club, the company is on track to spend an incredible $145bn (£107bn) on AI infrastructure by 2026. "Investors are nervous about the scale of the spending, but Meta is counting on the push leading to significant returns," she said. "It's building out AI infrastructure to supercharge its platforms, and if it works, it could make its advertising and commerce empire even more dominant." This massive investment comes as Meta AI has already reportedly reached a billion users across its family of apps, which includes Instagram, Facebook and Messenger.

The new incognito mode is currently being rolled out to users on WhatsApp and the standalone Meta AI app. It represents a major attempt by the tech giant to win user trust as it integrates AI more deeply into the daily lives of billions.