Meta won’t launch their AI models in the EU

Meta – parent of Facebook, Instagram, and WhatsApp –  will not launch its upcoming multimodal artificial intelligence (AI) models in the European Union (EU) due to the “unpredictable nature” of its regulatory environment.

“We will launch a multimodal Llama model in the coming months, but not in the EU due to the unpredictable nature of the European regulatory environment,” the company has advanced to Axios media.

This decision may have several consequences for both companies and users in the EU. One of them is that European companies will not have access to this multimodal AI technology to carry out their work processes, despite the fact that these are models that Meta always publishes under an open license.

In this line, users and companies will also not be able to develop their own products if they make use of the new models that Meta plans to launch. Similarly, the technology company has pointed out that it could also affect other companies that operate outside the EU and that, therefore, do use the multimodal models, as it could prevent these companies from offering their products and services in Europe.

Despite the fact that the Artificial Intelligence Regulation will come into force on August 1 this 2024, Meta is currently facing problems with the European regulation due to the need to use data from users participating in its platforms to continue training its AI models.

Specifically, the EU opposes the use of user data by companies such as Meta under the requirements of the current General Data Protection Regulation (GDPR), which seeks to ensure that users have greater control over the personal information handled by its online services, as well as deciding the type of advertising they receive.

Meta wants to use public info to train their AI 

In this context, Meta announced last May its intention to use information shared in Facebook and Instagram posts by public users to train its upcoming AI models. However, Meta later added an option so that, through an objection form, users could prevent using their data posted on both platforms.

Finally, in June, Ireland's data protection authority, the Data Protection Commission (DPC), announced that Meta had halted plans to train its AI models on public content shared by adults on Facebook and Instagram in the EU.

For its part, Meta shared its “disappointment” at having to delay training its large language models (LLMs) using public content from its platforms. “This is a step backward for European innovation, competition in AI development, and further delays in bringing the benefits of AI to people in Europe,” it sentenced in a statement on its blog, noting that “AI training is not a technology unique to Meta's services and that they are also more transparent than many of their industry peers.”

It is worth noting that other tech companies have also made this decision, as is the case of Apple, which recently detailed that it will delay the arrival of its AI features, Apple Intelligence, in the EU, as well as iPhone Mirroring for Mac and SharePlay Screen Sharing features due to “regulatory uncertainties” related, in this case, to the Digital Markets Act (DMA).

However, despite all this, Meta has also indicated to the media above outlet that they plan to release a larger version of their text-only Llama 3 model. This new version will be available to customers and companies in the EU. 

Likewise, the technology company has also confirmed that it will launch its following multimodal AI models in the United Kingdom because, although they also have a law similar to the GDPR, they do not have the same uncertainty scenario as in the European Union; it has clarified.