Meta's plan to expand its AI presence in South America has been hamstrung, with Brazilian officials today banning the Facebook parent company from using its citizens' data to train AI algorithms.

The National Data Protection Authority (ANPD) issued what it described today as a "preventative measure" that would immediately suspend changes made last month to Meta's privacy policy that added the company's AI training to its explanation of how it processes user data.

"After a preliminary analysis, given the risk of serious and difficult-to-repair harm to users, the Authority provisionally ordered the suspension of the privacy policy and the processing operation," the ANPD said in a notice we automatically translated from Portuguese to English.

ANPD took umbrage with what it said was Meta's "inadequate legal hypothesis for the processing of personal data," which Zuckerborg reportedly justified by saying it was in the legitimate interest of the company. "Such a hypothesis cannot be used when there is processing of sensitive personal data," the ANPD said. 

Brazilian officials were also upset about what they said was a lack of clear disclosure from Meta about privacy policy changes, too many limitations on the rights of users, and processing the data of minors "without due safeguards."

The ANPD said it was first clued into problems with Meta's processing of Brazilians' data due to suspected violations of the country's general data protection law, and made its determination after "sufficient preliminary findings" that Meta was in the wrong. 

Meta maintains that it's in the right.

"AI training is not unique to our services, and we're more transparent than many of our industry counterparts who have been using public content to train their models and products," a Meta spokesperson told The Register. "Our approach complies with privacy laws and regulations in Brazil, and we will continue to work with the ANPD to address their questions."

"This is a step backwards for innovation, competition in AI development and further delays bringing the benefits of AI to people in Brazil," the Meta spokesperson added. Brazil has one of the largest Facebook user bases in the world with an estimated 112 million on the platform as of April 2024, with only India, the United States, and Indonesia having more.

Meta caved to similar requests from EU officials in June, complaining that a pause on AI training on EU citizens' data would be to the detriment of Europeans too.

Meta first announced plans to expand availability of its AI products outside the US in April, and in May indicated it was going to start training its AI products on the data of Brazilian citizens. The company said then that its approach and opt-out options were "consistent with how we operate and how other technology companies are developing and improving their AI experiences."

Of course, that doesn't mean any of those companies are complying with Brazilian law so the investigation will continue. The preliminary order issued today, the ANPD said, was simply to stop Meta's AI data collection in Brazil before the country made a final determination in the matter.

Meta has five business days to comply with the order, or could face fines of R$50,000 per day (around $8,900 or £7,000) for noncompliance. Meta reported a net income of $12.3 billion in the first quarter of 2024 so it'll probably be able to afford it. ®

https://www.theregister.com//2024/07/03/brazil_meta_ai_training/