PARIS: Social media giant Meta announced on Monday that it will begin training its generative artificial intelligence (AI) models using public content and interactions from users in the European Union, resuming efforts previously paused due to privacy concerns.
According to a blog post by Meta, the company will use public posts and comments shared by adult users on its platforms — including Facebook and Instagram — across the 27 EU member states.
It will also collect data from user interactions with its Meta AI assistant, such as questions and queries, to help improve its AI systems.
The move follows Meta’s recent rollout of its AI assistant in Europe, months after it was launched in the United States and other major markets.
The company stated that the decision aims to ensure its AI models reflect regional linguistic and cultural nuances, including “dialects and colloquialisms, hyper-local knowledge, and the distinct ways different countries use humour and sarcasm.”
Users in the EU will receive notifications explaining the data use and will be given the opportunity to object through an online form.
“We’ll honour all objection forms,” Meta said in the post.
Meta emphasised that it will not use private messages for AI training.
It also positioned its decision as being in line with industry practices, noting that competitors such as Google and OpenAI have already trained their models on data from European users.
The company’s original plans were delayed last year after concerns were raised by privacy advocates, including the Vienna-based group NOYB (None of Your Business), led by data protection activist Max Schrems.
NOYB filed complaints with national privacy regulators, arguing that Meta’s proposed data use risked violating the EU’s strict General Data Protection Regulation (GDPR).
In its latest statement, Meta said that a panel of EU privacy regulators had “affirmed” in December that its revised approach complies with legal obligations under GDPR.
The company’s plans also follow a similar initiative launched last year in the United Kingdom, where Meta began training its AI models using public data from British users.
Like EU citizens, UK users are protected by strong data privacy laws that grant individuals control over how their personal data is collected and used.
While Meta’s announcement focuses on new data collection, the company had previously acknowledged that it had already used public text and images posted by adult Facebook users since 2007 to train earlier versions of its AI systems. – Agencies
