Meta has revealed plans to commence training artificial intelligence (AI) using data from EU users of its applications, including Facebook and Instagram. The data for this training will consist of publicly available posts, comments, and interactions with Meta’s AI, though it explicitly excludes private messages exchanged with friends and family. This initiative will only affect individuals over 18 years of age.
Meta has indicated that it will begin notifying EU users about the training process this week through in-app notifications and emails. These communications will include a link to a form allowing users to opt out. Users can also refer to Meta’s privacy policy, which notes that the company is still postponing the AI training plans based on feedback from regulators. This pause was initially requested by Irish regulators last year.
The company asserts that utilizing EU user data for AI training is aimed at developing models that accurately represent the unique cultural and linguistic characteristics of the region. This includes a focus on dialects, colloquialisms, and localized humor, which is particularly important for multi-modal AI that engages with text, voice, video, and imagery.
This announcement builds upon Meta’s earlier commitment to begin training its AI models with data from British users, who also enjoy heightened protections regarding personal data compared to US citizens. The data Meta plans to collect from EU users now seems minuscule when compared to the extensive dataset it disclosed last year, indicating that all public text and photos posted by adult Facebook users since 2007 had been used for AI training.
