Meta Halts AI Training in EU, Brazil Amid Regulatory Concerns

Meta's generative AI efforts have encountered regulatory challenges in the EU and Brazil as regulators have taken issue with the company’s data training practices.

Meta, the parent company of Facebook and Instagram, won't be introducing its multimodal AI models to European users, citing a lack of clarity around data rules.

The EU’s privacy watchdog requested the social media firm voluntarily pause training efforts of all future AI models using user data from European users.

Rob Sherman, Meta’s deputy privacy officer and vice president for policy, confirmed to FT that the European Data Protection Supervisor contacted the company and that it is complying with the request for now.

Meta will also withhold its next multimodal AI model, including future multimodal systems, from European products and services. Sherman said this could mean, however, that EU users could miss out on new AI features like in its Meta AI chatbot.

“If jurisdictions can’t regulate in a way that enables us to have clarity on what’s expected, then it’s going to be harder for us to offer the most advanced technologies in those places…  it is a realistic outcome that we’re worried about,” Sherman told the FT. 

The company has been working on a variety of multimodal systems for some time, including a version of its flagship Llama model capable of handling inputs beyond solely text, including video, images and audio.

Related:Meta to Use Public User Data for AI Training, Allows EU Opt Out

The multimodal models would also be featured in Meta’s Ray-Ban smart glasses, which it has been quietly improving for some time. The glasses feature a conversational AI assistant capable of generating information about what the wearer is observing.

However, Meta remains concerned about what it describes as an “unpredictable nature of the European regulatory environment.”

Meta is uncertain whether its use of European user data to train AI models complies with data protection laws, including the General Data Protection Regulation (GDPR), according to Axios.

The company reportedly briefed EU regulators on its AI plans some months ago but received only minimal feedback.

Meta is also concerned Europe’s strict data rules could impact businesses using its AI models, including that enterprises might not be able to license its multimodal systems.

There are also concerns that businesses outside the EU using Meta’s AI models may be unable to offer products and services in Europe that utilize the new multimodal models.

Meta offered a potential solution in June, allowing EU users to opt out of having their data used to train the company’s AI models.

Related:Meta Launches AI Assistant Across Facebook, Instagram, WhatsApp

Dieter Gerdemann, European lead for digital and analytics at management consulting firm Kearney, said the data protection supervisor’s request will be generally applicable to all technology companies using European-sourced data to train their AI models.

“It will force a break in current practices for AI model training, precipitating a dialogue on secure, future-ready legislation that regulates and guides the use of training data and its prerequisites,” Gerdemann said. “While the long-term solution is not yet clear, the call from the Data Protection Supervisor enforces clarity and action, ending a period of uncertainty for both consumers and technology companies. Will this change be written into law? That will depend on how the technology sector reacts to this temporary moratorium.”

Meta AI Chatbot Goes Dark in Brazil

Meanwhile, Meta has suspended its Meta AI chatbot service on its suite of social media apps in Brazil. 

The decision comes shortly after the country’s Data Protection Authority (ANPD) barred the company from using the personal data of Brazilian citizens to train its AI models.

Meta revised its privacy policy in May to include permissions for using user data and posts to train AI models. In June, it affirmed it wouldn’t be using private user messages or content from accounts of users under 18.

Related:Meta Unveils Largest Open-Source AI Model In History

The ANPD took issue with the policy change, suspending the change earlier this month. 

The agency was concerned that Meta failed to provide adequate and necessary information on what data was being used and the potential risk to younger users.

ANPD director Joacil Basilio Rael said the policy posed an “imminent risk of serious and irreparable damage” to Brazilians in a letter to Meta.

Meta has opted to suspend the chatbot service in Brazil while it engages in discussions with the ANPD, according to Reuters.