Brazil has blocked Meta from using Brazilian Instagram and Facebook posts to train its artificial intelligence (AI) models.
This comes weeks after the company abandoned plans to use posts from its UK and European users for similar purposes.
Brazil’s National Data Protection Agency (ANPD) said on Tuesday it would immediately suspend Meta’s updated privacy policy, which allows the company to train generative AI models such as chatbots based on user posts.
A Meta spokesperson told the BBC the company was “disappointed by the decision”, adding that its actions complied with local privacy laws.
“This is a setback for innovation and competition in AI development, further delaying the benefits of AI to the Brazilian people,” it added.
Meta has a big market in Brazil, which has 102 million Facebook users and more than 113 million Instagram users.
The ANPD said it acted because of “an imminent risk of serious and irreparable harm or the difficulty of restoring fundamental rights” for those affected. [account] “holder”.
Meta has five working days from the ANPD’s decision to show that it has amended its privacy policy to ensure that personal information found in public posts is not used to train its generative AI, or it faces a fine of 50,000 reais (£6,935) per day if it fails to do so.
The company’s updated policies have also come under scrutiny in the UK and the European Union.
The privacy policy changes, which are set to take effect in the region on June 26, will mean that Meta users’ information will be used to “develop and improve” the company’s AI products.
In Europe, the policy change applies to public posts, images, image captions, comments and stories made by users over the age of 18 on Facebook and Instagram, but does not include private messages.
However, these plans were put on hold after Meta announced that it had received a request from the Irish Data Protection Commission (DPC), on behalf of other European stakeholders, to postpone the training of its large language model (LLM).
LLM is the type of artificial intelligence that powers chatbots such as OpenAI’s ChatGPT and Google’s Gemini.
When announcing the delay on June 14, Meta said it was a “setback” for AI in Europe.
However, Meta decided to push for policy change in Brazil.
Pedro Martins of Data Privacy Brazil welcomed the ANPD’s decision, telling the BBC there was a discrepancy between Meta’s data protection measures for its Brazilian and European users.
Meta planned to use posts from children and teenagers in Brazil to train its AI models, but in Europe posts from people under the age of 18 would not be used, he said.
Brazil’s data protection regulator also found that personal data found in posts from children and teenagers could be collected and used to train Meta’s AI systems, potentially violating the country’s personal data protection laws.
Martins further noted that the steps users can take to stop Meta from using their personal information are easier in Europe than in Brazil, where it can take up to eight steps for a user to stop Meta from using their posts.
The BBC asked Meta to respond to allegations that the company planned to use posts from Brazilian children and teenagers to train its AI models, and whether it had imposed additional hurdles for Brazilian users to opt out.
