Meta AI, the artificial intelligence tool from Meta (owner of WhatsApp, Facebook, and Instagram), has launched in Brazil, sparking curiosity and concerns about user data privacy. Announced in September 2023, Meta AI aims to integrate AI functionalities across all Meta platforms, functioning similarly to a chatbot like ChatGPT.
Powered by Meta's Llama 3.2 language model, Meta AI utilizes a vast dataset comprising trillions of words from diverse sources, enabling it to provide relevant and reliable information. However, its reliance on user-generated content from Meta's platforms for training has raised eyebrows regarding compliance with Brazil's General Data Protection Law (LGPD).
After initial restrictions from the National Data Protection Authority (ANPD) regarding data collection for AI training, Meta developed a compliance plan that was approved in August 2024. This plan allows data collection only from public user posts on Facebook and Instagram, while the situation for WhatsApp is more complex due to its end-to-end encryption.
Meta assures users that the AI cannot read personal messages in WhatsApp, as it only processes messages that explicitly mention '@Meta AI'. Data collected for training is limited to user interactions with the AI and publicly available information. However, concerns remain about how user consent is obtained and whether all or part of the conversations are utilized for training.
Despite assurances, users cannot remove the Meta AI from WhatsApp, raising further privacy questions. Users can exercise their right to oppose data usage for AI training through a form provided by Meta, but the AI remains integrated into the app regardless of this choice. The ongoing debate highlights the tension between innovation in AI and the imperative of user privacy in an increasingly digital world.