Meta Directed to Cease AI Training Using Brazilian Personal Data

Brazil's data protection authority (ANPD) has prohibited Meta from using Brazilian personal data to train its artificial intelligence models due to concerns over potential harm to users. This decision follows an update to Meta's privacy policy in May, which allowed the company to utilize public data from platforms like Facebook, Messenger, and Instagram — including posts and images — for AI training.

The ruling stems from a Human Rights Watch report indicating that LAION-5B, a major image-caption dataset employed in AI training, includes identifiable photos of Brazilian children, heightening their risk of exploitation, including deepfakes. According to The Associated Press, ANPD expressed that the updated policy poses an “imminent risk of serious and irreparable damage” to the fundamental rights of Brazilian users. With 102 million users on Facebook in Brazil, the nation represents one of Meta's largest markets.

The ANPD has given Meta five working days to comply with this order, or the company will incur daily fines of 50,000 reais (about $8,808). In response, Meta stated that its policies align with Brazilian privacy laws and claimed that the ruling hinders innovation and competition in AI development, ultimately delaying benefits for Brazilian users. While Meta claims users can opt out of data use for AI training, ANPD argues that the process involves "excessive and unjustified obstacles."

Previously, Meta faced regulatory challenges in the EU, leading the company to pause its plans to train AI models on European Facebook and Instagram data. Meanwhile, the updated data collection policies remain in effect in the US, where user privacy protections are less stringent.