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

Meta has announced it will be using user data and posts to train its AI models — with only EU users able to opt out.

The company said it will use publicly available content from its social media platforms, such as Instagram and Facebook, to train its foundation models. However, it will not use private user messages or content from accounts of users under 18.

Meta said it needs to train its models on public content otherwise its AI “won’t accurately understand important regional languages, cultures or trending topics on social media.”

Currently, there are no explicit rules in the EU that prohibit companies from using user data to train AI models. However, EU data protection laws, specifically the GDPR, require businesses to obtain explicit consent from users before using their data. Businesses are also obligated to be transparent about how they’re using user data, providing the option to withdraw consent at any time.

Meta says if Europeans opt out, they will be “ill-served by AI models that are not informed by Europe’s rich cultural, social and historical contributions.”

Other companies developing foundation AI are doing the same, including OpenAI and Google, however, Meta claims it’s being more transparent than its rivals. The company says it has sent billions of notifications and emails to European users informing them they can opt out.

Related:Meta’s Chameleon AI Model Seamlessly Handles Text and Images

“Our approach is more transparent and offers easier controls than many of our industry counterparts already training their models on similar publicly available information,” Stefano Fratta, Meta’s global engagement director for privacy policy wrote in a blog post.

Meta didn’t offer the opt-out policy for Llama 3 as the model had already finished development, however, the recent announcement applies to its next line of foundation models.

“Models may be trained on people’s publicly shared posts, but it’s not a database of each person’s information nor is it designed to identify any individual,” Fratta said. “Rather, these models are built by looking at people’s information to identify patterns, like understanding colloquial phrases or local references, not to identify a specific person or their information.”

Meta’s transparency announcement comes as the company was the subject of 11 complaints by a nonprofit over its data handling practices related to AI.

Noyb filed the complaints, asking several EU Member States to launch an urgency procedure available through the GDPR to stop the company from using user data before the June 26 deadline.

Related:Meta Unveils Llama 3, the Most Powerful Open Source Model Yet

The nonprofit claims that Meta will use all public user data that it has collected since 2007  for its future AI efforts.

"Meta is basically saying that it can use ‘any data from any source for any purpose and make it available to anyone in the world’, as long as it’s done via 'AI technology.' This is clearly the opposite of GDPR compliance,” said Max Schrems, noyb’s chairman and founder. “‘AI technology’ is an extremely broad term. Much like ‘using your data in databases,’ it has no real legal limit. Meta doesn't say what it will use the data for, so it could either be a simple chatbot, extremely aggressive personalised advertising or even a killer drone. Meta also says that user data can be made available to any ‘third party’ — which means anyone in the world."