Meta Platforms announced on Wednesday a new privacy feature for its artificial intelligence assistant, Meta AI, on the WhatsApp messaging application. The feature, termed "incognito mode," allows users to engage in temporary, private conversations with the AI chatbot.
Feature Details
The incognito mode processes messages in a secure environment that Meta cannot access. Conversations conducted in this mode are end-to-end encrypted and are not logged or stored on Meta's servers. Messages disappear automatically when the user exits the session.
Meta CEO Mark Zuckerberg stated that no record of conversations is kept on the company's servers. The feature currently supports text inputs and outputs only. Image uploads and AI image generation are not available in this mode. Users must verify they are at least 13 years old to use the feature.
Technical Implementation
Meta published a technical white paper describing the infrastructure for private processing. The system utilizes servers powered by AMD and Nvidia processors and employs anonymizing routing to prevent identification of user queries.
Future Developments
WhatsApp plans to introduce a feature called "Side Chat," which would enable private AI assistance within ongoing conversations without interrupting the main chat thread.
Context and Limitations
The feature addresses privacy concerns related to generative AI systems, where large language models are trained on data that may include user interactions. Other AI chatbots, including Google's Gemini and OpenAI's ChatGPT, already offer similar privacy controls, such as options to disable chat history or opt out of data use for model training.
"Users are asking significant personal questions of AI systems and should not have to share that information with the companies operating the systems." — Will Cathcart, Meta's head of WhatsApp
He noted the incognito mode includes safety features designed to refuse harmful responses and may cease interaction when necessary.