These models, which can reason across video, audio, images, and text, will be incorporated into products like smartphones and Meta Ray-Ban smart glasses. European companies will not be able to use these models, even under an open license, and companies outside the EU may also be affected.

Meta plans to release a larger, text-only version of its Llama 3 model for EU customers. The issue lies with GDPR compliance, not the AI Act.

Meta intended to use public posts from Facebook and Instagram for training, notifying over 2 billion EU users with an opt-out option. Despite briefing EU regulators in advance, Meta was ordered to pause training on EU data in June.

The UK, with similar laws to GDPR, does not face the same regulatory uncertainty, and Meta plans to launch its new model there. A Meta representative noted that European regulators take longer to interpret existing laws. Training on European data is crucial for products to reflect regional terminology and culture.