- Citing regulatory unpredictability, Meta will not introduce its multimodal AI models in the European Union.
- The decision reflects broader compliance challenges with EU data protection laws, particularly the General Data Protection Regulation (GDPR).
OUR TAKE
Meta has decided to suspend the launch of its latest multimodal AI model in the EU due to uncertainty about the regulatory environment in the region. This decision will not only have an impact on Meta’s own technology release plans, but could also have a knock-on effect on the development of the entire AI industry in Europe. Against this backdrop, Meta’s choice has triggered deeper thinking in the industry about the balance between technology ethics, data privacy protection and freedom of innovation.
–Elodie Qian, BTW reporter
What happened
Meta has announced that it will not be offering its forthcoming multimodal AI models in the European Union, citing regulatory uncertainty.
The EU recently set a compliance deadline for the new Artificial Intelligence Act, requiring tech companies to comply by August 2026. Meta’s decision is similar to that of Apple, which recently signalled it may exclude the EU from its Apple Intelligence rollout due to concerns over the Digital Marketplace Act.
The multimodal models, designed to process text, images, and audio, are integral to Meta’s platforms and its Ray-Ban smart glasses. Meta said, “We will release a multimodal Llama model over the coming months, but not in the EU due to the unpredictable nature of the European regulatory environment.”
The EU’s competition commissioner, Margrethe Vestager, criticised Apple’s decision, suggesting it was a move to disable competition. Meta’s withholding of its AI models could impact companies building products and services on these models, as they would be unable to offer them in Europe.
An EU spokesperson, Thomas Regnier, said the regulator does not comment on individual company decisions but reiterated that all companies are welcome to offer services in Europe as long as they comply with the bloc’s laws, including the upcoming Artificial Intelligence Act.
Also read: Meta releases early versions of Llama 3 multimodal AI model
Also read: ‘EU AI Act’ takes effect in August: A landmark regulation for AI
Why it’s important
Despite its concerns, Meta plans to release its text-only model, Llama 3, in the EU. The company’s primary concern is training AI models with data from European customers while adhering to the General Data Protection Regulation (GDPR). Meta had to cease using publicly available posts from Facebook and Instagram users in the EU for AI model training after facing resistance from data privacy regulators.
Meta argued in a blog post that training models on public content shared by Europeans is essential to accurately understand regional languages, cultures, and social media trends. The company believes that without this cultural input, AI models will not serve Europeans well.
Although Meta is cautious about releasing multimodal models in the EU, it plans to launch them in the UK, which has similar data protection laws. The company noted that European regulators are slower in interpreting existing laws compared to other regions.
Meta’s move highlights a growing conflict between the US tech giant and European regulators. The tensions are nothing new, as the EU has long been seen as regulating privacy and antitrust matters more tightly. In turn, tech companies argue that these regulations hurt both consumers and the competitiveness of European companies.
