WhatsApp has launched a feature to keep conversations with AI (artificial intelligence) chatbots completely private. With this new ‘Incognito’ mode facility, the messaging platform itself will not be able to read any of the user’s messages.
That is, when this feature is turned on, none of the replies given by the user or the AI chatbot will be monitored. Also, old conversations will be automatically deleted from the user’s chat box.
Will Cathcart, head of WhatsApp, said that people want to have personal conversations with AI about sensitive topics like health, relationships or money that no one else can see. This step has been taken considering the needs of the people.
However, when talking to the BBC, a cyber security expert warned that since Meta or WhatsApp do not have these chat histories or information, it will be difficult to ensure the accountability of the organization if there is a problem.
Other popular platforms owned by Meta are Instagram, Facebook and Messenger. When ‘Meta AI’ was added to WhatsApp last year, many users were outraged that there was no option to turn it off. However, in May 2025, Mark Zuckerberg, the head of Meta, said that the users of Meta AI have exceeded 100 million in their various apps.
Cathcart said of the new feature, “We’ve heard from many people that they are uncomfortable sharing personal information with organizations, but they also want to get the answers they need.”
Currently, most AI organizations store chatbot usage data. Even data from normal users without premium enterprise accounts is used to train future AI models.
Mark Zuckerberg described the initiative as the first major AI product where no records or ‘logs’ of conversations would be stored on servers.
Will Cathcart also said that WhatsApp’s incognito mode and the ‘end-to-end encryption’ technology used in normal messages are not the same, but will provide equal protection.
Professor Alan Woodward, a cyber security expert at the University of Surrey, believes that the introduction of a second system is less likely to disrupt WhatsApp’s existing security measures. However, he expressed concern that due to the incognito mode, errors or misuse of the AI could be hidden.
Several AI firms, including OpenAI and Google, have already faced wrongful death lawsuits. Professor Woodward said there was a risk that AI would lack accountability for wrong answers.
He told the BBC: ‘Personally, I think AI should be privy to what someone wants to know. Because people ask many personal questions. But you have to have a lot of faith that the AI won’t mislead the user.’
The main concern is that deleted messages cannot be recovered by either the user or the meta. As a result, if a conversation leads to someone’s harm, death or suicide, it will be impossible to prove it.
Cathcart said Incognito mode will initially only work with text, not images. Meta AI’s security system or ‘guardrails’ will be very careful not to answer any questions that may appear harmful or illegal.
With WhatsApp shutting down any other chatbots, billions of users are only getting access to Meta’s own AI on the platform.
Susannah Streeter of investment platform Wealth Club said: ‘Meta is on track to spend $14.5bn (£10.7bn) on AI infrastructure by 2026. Investors now want to see huge returns against this huge investment.’
He further said, ‘Even though investors are worried about such huge expenditure, Meta expects it to bring big profits in the future. They are building AI infrastructure to power their platforms. If it succeeds, their advertising and trading empire will become even more monopolistic.’
