Meta will start training its AI models using everyone's social media posts though European Union users can opt out, a luxury the rest of the world won't enjoy.

The move, which the Facebook parent detailed in an announcement today, is ostensibly to bring its machine-learning systems to Europe.

Meta has so far not included its European userbase in its AI training data, presumably to avoid legal conflict with the continent's privacy regulations. Now it's pushing ahead with that despite complaints.

"To properly serve our European communities, the models that power AI at Meta need to be trained on relevant information that reflects the diverse languages, geography and cultural references of the people in Europe who will use them," the social media titan said.

"To do this, we want to train our large language models that power AI features using the content that people in the EU have chosen to share publicly on Meta's products and services."

As training AI from user data is doubtlessly going to be contentious in Europe, Meta has attempted to cover itself in two ways. Firstly, when it says "public content," Meta means posts, comments, photos, and other content posted on its social media platforms by users over the age of 18. Private messages are, apparently, strictly verboten from the training data.

Meta also says it has sent billions of notifications to European users since May 22 to give them a chance to decline before the AI training rules kick in worldwide on June 26. The Instagram goliath says any user can decline, no questions asked, and that their posts won't be used to train AI models now or ever.

This is substantially different from the rest of the world, where opting out just isn't a choice. Granted, it's already too late to opt out for training data used for Meta's LLaMa 3, but even training for future models is mandatory for Facebook and Instagram users outside of the EU. Perhaps users outside of Europe will be able to choose to opt out in the future, but for now it's a feature exclusive to the EU.

Although Meta likely feels that it's in a good position to start using European user data, it's hard to imagine there being no pushback at all. Before the social media giant even made its public announcement, it signaled its intentions via an update to its privacy policy last week. That prompted consumer privacy advocacy group noyb to file complaints across Europe.

Noyb claims the collection of user data needs to be opt-in, not opt-out, by default. The fact that data can't really be scrubbed from an LLM or other AI model is also likely to cause problems due to the European Union's Right to be Forgotten.

Plus, Meta and the EU are not on the best of terms. Just this year the EU launched probes into Meta including one concerning child safety and another about misinformation surrounding the now-concluded EU parliamentary elections. While it's not clear if Meta will have its way in the end, it's hard to imagine there not being a challenge against the social network at some point or another. ®

https://www.theregister.com//2024/06/10/meta_ai_training/