After stopping last year, Meta will soon begin training its AI models in Europe based on public posts and comments from adult users. The goal is to teach AI to "better understand and reflect cultures, languages, and histories" to "better support millions of people and businesses in Europe," Mark Zuckerberg's company emphasizes.
You can opt out by filling out a form. Starting this week, users in the European Union who use Meta platforms will begin receiving notifications, both in apps and by email, explaining the type of data the company will begin using. Also included is a link to a form where you can opt out “at any time of having your data used in this way.” The AI will be trained using public content from adult users only, i.e. public posts and comments, not private messages exchanged with friends and family. User interactions with Meta AI will also be used, such as questions and requests made on Instagram and Messenger, not those on WhatsApp. The smart chat feature marked by the blue circle icon — which cannot be removed at the moment — was introduced last month in Europe after Meta stopped its launch last year due to concerns from the Irish data protection authority.
The company thus recalls that in 2024 it had postponed the training of large language models using public content, "waiting for regulators to clarify the legal requirements". "We regret that it took us almost a year to get to this point - notes Meta - We welcome the opinion provided by the European Data Protection Board in December which confirmed that our initial approach was compliant with our legal obligations. We have since engaged constructively with the Irish Data Protection Commission and look forward to continuing to bring all the benefits of generative AI to people in Europe".
“It is important to note that the type of AI training we are doing is not unique to Meta, nor will it be unique to Europe. This is how we have trained our generative AI models for other regions of the world since launch. We are following the example of other companies, including Google and OpenAI, who have already used European user data to train their AI models,” the company added. “It is crucial that our generative AI models are trained on This is particularly important at a time when AI models are becoming increasingly advanced with multimodal capabilities that include text, voice, video and images.”
