Meta to Train AI Models Using Public EU User Content
Meta announced that it will begin training its generative AI models using public content, including public posts and comments shared by adult users in the European Union. The initiative also includes using interactions with Meta AI, such as questions and queries, to improve its capabilities. The move follows Meta AI’s recent launch in Europe and aims to better tailor AI systems to the languages, cultures, and local nuances of EU users.
User Notifications and Opt-Out Form
Starting this week, Meta users in the EU will begin receiving in-app and email notifications about how their public content may be used for AI training. These alerts will also link to an easy-to-use opt-out form, allowing users to object to their data being included in training. Meta has pledged to honor all opt-outs, including those already submitted.
The company clarified that it does not use private messages or data from underage accounts for this purpose.
Why Meta Is Taking This Step
Meta explained that to build AI that truly serves and reflects European users, the training data must include the rich diversity of public interactions across its platforms. This includes regional languages, dialects, humor, and other unique cultural elements, especially as the company advances into multi-modal AI, covering text, voice, video, and imagery.
Meta emphasized that similar AI training practices are already in use by competitors such as Google and OpenAI, and that their approach in Europe is more transparent than many others in the industry.
Also read: Ireland's DPC Investigates X Over AI Training with EU User Data
Regulatory Compliance and Legal Backing
After pausing plans last year pending regulatory clarity, Meta says its current approach aligns with European legal standards. The company cited favorable guidance from the European Data Protection Board (EDPB) and its ongoing collaboration with the Irish Data Protection Commission (IDPC) as confirmation that its practices meet legal obligations.
