WhatsApp has taken a new step to protect user privacy. The company has added a new mode to its platform to secure conversations between users and its AI chatbot.
The special feature of this new system is that even Meta itself will not be able to read chats between users and the AI. The feature works through a new “Incognito” mode. This means that when the feature is enabled, conversations between the user and the AI will not be monitored, and all previous chats will automatically disappear once the conversation ends.
Will Cathcart, head of WhatsApp, said that people want to talk privately with AI about sensitive topics such as health, personal relationships, or finances, and they do not want others to see that information. However, a cybersecurity expert warned that if chat histories cannot be reviewed, accountability could become weaker if any harmful incident occurs in the future.
After Meta added “Meta AI” to WhatsApp last year, many users expressed frustration because there was no way to disable it. However, in May 2025, Mark Zuckerberg said that the number of users of Meta’s AI across all its apps had already exceeded one billion. In a recent announcement, Cathcart said that many people feel uncomfortable sharing personal information with companies, while still wanting answers from AI.
Currently, most AI companies store chatbot conversation data. Unless users have premium enterprise accounts, those conversations are often used to train future AI models. Zuckerberg described this new feature as the “first major AI product” where no records or logs of conversations are stored on servers.
Cathcart explained that WhatsApp’s “Incognito” mode is not exactly the same as the platform’s standard end-to-end encryption technology, but it offers a similar level of effectiveness. University of Surrey cybersecurity expert Professor Alan Woodward said the new system reduces the risk of damaging WhatsApp’s existing security measures. However, he also expressed concern that the Incognito mode could hide AI misuse or inappropriate behavior.
Several AI companies, including OpenAI and Google, are currently facing lawsuits related to “wrongful death” cases. Professor Woodward said the Incognito mode may create a lack of accountability regarding AI-generated responses. He added that people should be able to keep their AI questions private because they often discuss deeply personal matters with AI systems. However, this also requires a high level of trust that the AI will not mislead users.
The main concern is that if messages are automatically deleted and cannot be recovered by either the user or Meta, it may become impossible to prove whether a conversation encouraged someone toward harm, death, or suicide. Cathcart stated that the Incognito mode will initially work only for text messages, not images. He also said that Meta AI’s safety systems will continue blocking responses to harmful or illegal requests.
WhatsApp has also restricted the use of other AI chatbots on its platform, meaning billions of users can now only access Meta’s own AI services within the app. Wealth Club representative Susannah Streeter said that Meta plans to spend $145 billion on AI infrastructure by 2026. Naturally, investors are expecting large returns from such massive investments.
She added that while investors are somewhat concerned about the huge expenses, Meta remains optimistic that the initiative will generate strong returns. The company is building AI infrastructure to strengthen its own platforms, which could make its advertising and commerce empire even more powerful if successful.
Comments
Post a Comment