Social media giant Meta is preparing to resume its use of EU citizens’ data to train its AI models nearly a year after pausing the plans following EU regulators’ concerns over data privacy and transparency.

The company said it will begin training its AI models on the public content shared by adults across its platforms, including posts and comments, as well as the interactions people have with Meta AI, the chat function launched across the EU last month.

By using the troves of data spanning Facebook, Instagram, WhatsApp and Messenger, Meta said its AI could be taught to ‘better understand’ and reflect EU countries diverse cultures, languages and history, claiming this would help the company support millions of people and businesses across the continent.

“We believe we have a responsibility to build AI that’s not just available to Europeans, but is actually built for them,” said Meta in a blog post describing its plans. 

“That’s why it’s so important for our generative AI models to be trained on a variety of data so they can understand the incredible and diverse nuances and complexities that make up European communities. 

“That means everything from dialects and colloquialisms, to hyper-local knowledge and the distinct ways different countries use humour and sarcasm on our products.”

Meta said that beginning this week, its users will start receiving notifications, via app and email, to explain the data use, including a clear opt-out link for those who don’t want their data used in training the firm’s AI. Meta also committed to honouring all the objection forms it had already received.

Although Meta is already using user-generated data to train AI in the US, Europe’s GDPR framework has proved a significant barrier to using that same approach across the Atlantic, limiting the firm’s efforts to refine its LLMs on EU-based data.

Last year, the company put plans for training its large language models using public content on ice following a request from the Irish Data Protection Commission (DPC), calling it a ‘step backwards’ for European innovation, though also faced backlash from advocacy group NOYB (none of your business), which urged the EU’s other national privacy watchdogs to stop Meta from using years worth of user data for ‘undefined’ AI development.

Recommended reading

• Landmark Meta Antitrust Case Is A True Test of the FTC’s New Direction
• Meta Ends Fact-checking Citing Censorship Concerns
• Tech Giants Reconsider DEI Ahead of Trump Inauguration

With this turnaround, Meta stressed it has taken pains to address these concerns, citing an opinion from the European Data Protection Board in December, which provided some guidance for the use of data by firms developing AI models.

“Since then, we have engaged constructively with the IDPC and look forward to continuing to bring the full benefits of generative AI to people in Europe,” said Meta.

Meta also pointed out that it is now far from the only big tech firm taking part in this kind of AI training, saying that it is ‘following the example set by others, including Google and OpenAI’, both of which have already used data from European users to train their AI models.

Despite Meta’s attempts to be transparent, there are still plenty of questions being asked about how tech firm’s AI training fits beside EU data laws, with the Irish data regulator only last week opening a new inquiry into the use of personal data by X in training its genAI model, Grok.