Meta Platforms announced the rollout of an “incognito” mode for WhatsApp users, enabling private conversations with its AI chatbot. This new feature aims to alleviate privacy concerns associated with users sharing sensitive information within their chats. According to a company blog post, the incognito chat mode facilitates private, temporary interactions with Meta AI, the artificial intelligence assistant that has been integrated into WhatsApp for several years.

Meta emphasized that messages within this mode will be processed in a “secure environment” inaccessible even to Meta itself. Furthermore, these conversations will not be saved by default and are designed to disappear automatically upon exiting a session. The introduction of such privacy features comes as generative AI systems, which rely on large language models trained on vast datasets often including personal user information, have faced scrutiny over data privacy.

Rival chatbot providers, such as Google’s Gemini and OpenAI’s ChatGPT, already offer comparable privacy functionalities. Google’s Gemini, for instance, allows users to disable chat history and opt out of contributing their data for AI model training. ChatGPT provides similar controls, empowering users to manage their data privacy.

Meta explained that the decision to implement incognito chats stems from the observation that users frequently pose sensitive questions to chatbots, often containing private financial, personal, health, or work-related data. Will Cathcart, Meta’s head of WhatsApp, stated that users are increasingly asking profound questions about their lives to AI systems and should not feel compelled to share such information with the companies operating these systems.

In addition to privacy, the incognito chat mode incorporates safety features. Cathcart detailed that the chatbot is designed to prevent responses to harmful topics, instead guiding users towards helpful information when possible. In cases where a topic is deemed inappropriate, the chatbot will refuse to answer and may eventually cease interaction with the user entirely. Users in this mode will be limited to typing questions and receiving text-based responses, with no capability to upload or generate images. Moreover, age verification will be required, as Meta prohibits users under the age of 13 from using its platforms.