Meta seeks to train AI model on European data as it faces privacy concerns
Meta, the parent company of Facebook, Instagram, and WhatsApp, has announced its plans to utilize data from users in privacy-conscious Europe to enhance its artificial intelligence models. As the social media giant competes with rivals like OpenAI and Google, it aims to address concerns related to data protection.
The company highlighted the importance of incorporating the "languages, geography, and cultural references" of European users into its AI models. To achieve this, Meta intends to leverage public data from European users to train its Llama AI large language model.
Challenges with European Data Privacy Laws
Meta's AI training efforts in Europe face challenges due to strict data privacy laws imposed by the European Union. These laws empower individuals to control how their personal information is utilized, posing obstacles for Meta's AI development endeavors.
Last week, Vienna-based group NOYB, led by activist Max Schrems, raised concerns with 11 national privacy watchdogs regarding Meta's AI training plans. The group urged authorities to intervene and halt Meta's initiatives before advancing with the training of Llama's next generation.
Training AI Language Models
AI language models rely on extensive datasets to enhance their predictive capabilities, enabling them to anticipate the most probable next word in a given sentence. With each new iteration, these models become increasingly sophisticated and proficient compared to their predecessors.
Meta's AI assistant feature is already integrated into Facebook, Instagram, and WhatsApp for users in various countries, excluding Europe. Meta emphasized the necessity of training its models on public content shared by Europeans to ensure accurate understanding of regional languages, cultures, and current social media trends.
Respecting User Privacy
Meta assured that it would not utilize private messages or content from European users under 18 years old for AI training purposes. The company has actively communicated its plans to European users through notifications and emails, offering the option to opt out through an online form.
Furthermore, Meta's updated privacy policy, effective from June 26, indicates the initiation of training for the next model shortly thereafter.
