Meta is facing increasing regulatory roadblocks in Brazil and the EU
Meta faces significant regulatory challenges in its AI initiatives, suspending its AI assistant in Brazil due to data protection concerns and deciding not to release its advanced multimodal Llama model in the EU because of an unpredictable regulatory environment.
Meta confirmed this week that it would pause AI training in Brazil and not release its most recent multimodal AI model in the European Union. In Brazil, the company has suspended the availability of its generative AI tools following objections from the National Data Protection Authority (ANPD) Meta's privacy policy provisions concerning personal data and AI. The ANPD cited "imminent risk of serious harm" to fundamental rights as the reason for this preventive measure; and ruled that Meta needed to exclude the section where it permits itself to train AI systems with users' public data.
The ANPD granted Meta 5 days to comply with its request, or the company could face daily fines of up to 50,000 reais (around $8,808). Although the ANPD recognizes that users can opt out of having their information used for training, the regulator found the process is unnecessarily and unjustifiably complicated. Meta confirmed that it would pause its generative AI services while it addresses any questions the ANPD may have. Brazil is one of the largest markets for Meta, with 102 million Facebook users in the country, and the second-largest WhatsApp user base after India.
Simultaneously, Meta has decided not to release an advanced multimodal version of its Llama AI model in the European Union. The company attributes this decision to the "unpredictable nature of the European regulatory environment," particularly concerning GDPR compliance and the upcoming EU AI Act. Meta was recently forced to stop AI training on European users' data for concerns similar to the ones expressed by the ANPD. Text-only versions of Llama are available in the European Union, and a text-only version of the newest Llama model that has not been trained on European user data will be launched soon in the region.
