- tech360.tv
Meta to Use European Public Data for AI Training Amid Privacy Concerns and EU Regulations Compliance
Meta intends to train its AI model on European data to better serve its European user base. European data privacy rules present obstacles for Meta's AI training efforts. The privacy activist group NOYB has filed a complaint against Meta's AI training plans.
The decision comes as Meta faces growing concerns about data security and tries to keep up with competitors such as OpenAI and Google.
Meta feels that in order to better cater to its European user base's languages, geography, and cultural references, public data from these users should be included in the training of its Llama AI large language model.
However, Meta's AI training efforts are hampered by the European Union's (EU) strict data privacy rules, which give individuals discretion over the use of their personal information. Activist Max Schrems' Vienna-based organisation, NOYB, has filed objections with 11 national privacy watchdogs, requesting them to intervene and block Meta's AI training plans for Llama's next generation.
AI language models use large datasets to anticipate the most likely next word in a sentence, with subsequent versions getting more complex. While Meta's AI assistant has been integrated into Facebook, Instagram, and WhatsApp for users in the United States and 13 other countries, it is not available in Europe.
Stefano Fratta, Meta's worldwide engagement director of privacy policy, underlined the need of training AI models with public content posted by Europeans on social media platforms. He added that without this training, the models would be unable to effectively understand regional languages, cultures, or popular themes. Fratta also stressed the importance of adding Europe's diverse cultural, social, and historical contributions into AI models.
Fratta said that other businesses, including Google and OpenAI, have already trained their models on European data. Meta promises users that it will not use any private communications or information from European users under the age of 18.
To address privacy concerns, Meta has issued 2 billion notifications and emails to European customers since May 22, outlining its goals and offering an online opt-out option.
The amended privacy policy, which goes into effect on June 26, signals that training for the next AI model will begin shortly thereafter.
Meta aims to train its AI model on European data to better cater to its European user base.
European data privacy laws pose challenges for Meta's AI training efforts.
Privacy activist group NOYB has filed complaints against Meta's AI training plans.
Source: AP NEWS