Meta has announced that it’s preparing to train AI on the data of EU users of its apps, including Facebook and Instagram. The company says that includes things like public posts, comments, and their chat history with Meta AI, but won’t include “private messages with friends and family.” It also only applies to those who are over 18, the company says.

According to Meta, it will start notifying its EU users about the training this week, via in-app notifications and email, and will include a link to an objection form for those who want to opt out. You should be able to find such a link in its privacy policy, which says as of this writing that, based on regulator feedback, the company is still delaying its plans to train AI models on EU user data. Meta put its AI-training plans in Europe on hold last year after being asked to do so by Irish regulators.

Meta claims it’s training AI on EU user data to help it create models that reflect the regions they’re being used in, including “everything from dialects and colloquialisms, to hyper-local knowledge and the distinct ways different countries use humor and sarcasm.” The company adds that this matters particularly for the text, voice, video, and imagery produced by multi-modal AI.

This follows Meta’s announcement last year that it would start training its AI models on British users, who, like those in the EU, are guaranteed more protections regarding the use of their personal data online than people in the US. What Meta will get out of users now is a pittance compared to what it may already have — the company admitted last year it had trained AI using all text and photos adult Facebook users had publicly posted since 2007.