Meta's Plan to Use EU User Data for AI Training Raises Privacy Concerns

AIbase基地

Published inAI News · 4 min read · Apr 15, 2025

Meta Platforms, Inc. announced its plan to use user data from its applications in the European Union (including Facebook and Instagram) to train artificial intelligence models. The company clarified that the training data will include public posts, comments, and chat history with Meta AI, but not "private messages with friends and family," and that training will only apply to users over 18 years old.

Meta, metaverse, Facebook

Meta stated that it will inform its EU users of this plan this week through in-app notifications and emails, providing a link to an opt-out form for users who wish to decline. Users can also find the link in Meta's privacy policy. It's noteworthy that, at the time of this writing, the policy indicates Meta is still delaying the plan to use EU user data for AI model training, pending regulatory response. Previously, Meta had suspended its AI training plan in Europe at the request of the Irish regulator.

Meta claims that using EU user data for AI training aims to create models that better reflect the region's users, including understanding "everything from dialects and colloquialisms to hyperlocal knowledge and the unique ways different countries use humor and irony." The company added that this is especially important for generating multimodal AI content, such as text, voice, video, and images.

Previously, Meta announced last year that it would begin using user data from the UK to train its AI models. Similar to EU users, UK users enjoy stronger protections regarding the online use of their personal data compared to US users. Meta acknowledged last year that its AI model training data came from publicly posted texts and photos from all adult Facebook users since 2007, and that the amount of EU user data used in this plan might be relatively small. This launch of Meta's AI training plan in the EU, and whether users can effectively exercise their right to opt out, will likely generate further debate on user privacy and data usage.

Meta Restarts AI Training Using Public Content from European Users

Meta recently announced it will resume training its AI models using publicly available content from European users. This decision follows a pause last year due to data privacy concerns. Meta stated that this AI training will primarily rely on publicly shared posts and comments from adult users across the 27 EU countries. Furthermore, interactions between users and Meta AI, such as questions and queries, will also be used to train and improve its AI models. Image attribution: Image generated by AI, image licensing provided by Midj

Meta's Llama-4-Maverick Plummets in Rankings, Raising Concerns of Benchmark Manipulation

Meta's open-source large language model, Llama-4-Maverick, has experienced a dramatic drop in LMArena rankings, plummeting from second place to 32nd. This significant shift has sparked widespread skepticism among developers, who suspect Meta may have manipulated the benchmark by submitting a specially optimized version. The issue stems from Meta's April 6th release of its latest large language model, Llama 4, encompassing three versions: Scout, Maverick, and Behemoth.

Musk's X Company Under Investigation for Allegedly Using User Data to Train AI Chatbot Grok

Ireland's Data Protection Commission (DPC) has launched a significant investigation into Elon Musk's social media platform, X. The investigation centers on whether X used European users' personal data to train its AI chatbot, Grok, without their consent. Under the EU's General Data Protection Regulation (GDPR), companies must obtain explicit consent before processing user data; violations can result in substantial fines. The DPC has stated that a thorough examination will be conducted.

Llama 4 Arrives on Vertex AI: Deploy Meta's New Model with One Click

Google Cloud Platform recently announced that Meta's latest generation of open-source large language models, Llama 4, is now available in its Vertex AI Model Garden. The news has generated significant excitement in the global tech community. The Scout and Maverick models from the Llama 4 series are now integrated into Vertex AI and available to developers via fully managed Model-as-a-Service (MaaS) API endpoints in preview.

Authors Protest Meta's Alleged Use of Their Works to Train AI

Last week, around 100 authors protested outside Meta's London headquarters, accusing the American tech giant of using their works without permission to train its AI models. Protesters chanted "Meta, Meta, book thieves," with some holding signs such as "I wanted to write a slogan, but you'll steal it" and "Give us back our books, Zuck," a clear jab at Meta CEO Mark Zuckerberg. Earlier reports suggested Zuckerberg approved the company's use of a dataset called

DeepSeek and Tsinghua University Collaborate on Self-Optimizing AI Model

Amidst the growing prevalence of artificial intelligence, the collaboration between DeepSeek and Tsinghua University has garnered significant industry attention. DeepSeek, a Chinese startup, is renowned for its breakthroughs in low-cost inference models. This collaboration aims to further reduce the training costs of AI models, thereby enhancing operational efficiency. DeepSeek recently launched a new low-cost inference model that has generated considerable market excitement. To further optimize this model, DeepSeek's research team...