Regional ambitions for multimodal Llama are abandoned by Facebook’s owner due to unpredictable privacy restrictions
Due to the “unpredictable” behaviour of regulators, Mark Zuckerberg’s Meta has opted not to release an advanced version of its AI model in the EU.
The company behind Facebook, Instagram, and WhatsApp is preparing to introduce the multimodal version of its Llama model, which will let it to function with text, video, photos, and audio instead of just one medium. Since Llama is an open-source model, users are allowed to download and alter it as they see fit.
A Meta representative did, however, clarify that the model will not be sold in the European Union. This ruling underscores the continuous conflicts in an increasingly restrictive regulatory environment between big tech and Brussels.
“We will release a multimodal Llama model in the coming months, but not in the EU due to the unpredictable nature of the European regulatory environment,” the representative stated.
The EU AI Act is being utilised in Brussels, and it will go into force next month. The Digital Markets Act (DMA) applies new regulations on large IT companies.
Regarding the multimodal Llama model, Meta made a decision based on GDPR (General Data Protection Regulation) compliance. Because of possible privacy regulatory violations, Meta has been told to stop using postings from Facebook and Instagram users in the EU to train its AI models.
The Data Protection Commission of Ireland, which is in charge of monitoring Meta’s adherence to the GDPR, announced that the firm and it are still in talks over model training.
Meta is concerned, meanwhile, that additional EU data watchdogs could interfere with the regulatory procedure and even put an end to approval. In the EU, text-based versions of Llama are available, and a new text-only version is on the way. EU Meta user data was not used in the training of these models.
This move follows Apple’s announcement last month that it would not be releasing some of its new AI features in the EU because of DMA compliance issues.
Meta has planned to incorporate its multimodal Llama model into gadgets such as cellphones and smart glasses from Ray-Ban. Axios published the first report of the Llama decision.
The government’s privacy worries over the usage of user data for model training prompted Meta to announce on Wednesday that it has stopped using generative AI tools in Brazil. According to the company, it will put these tools on hold until it has talks with Brazil’s data authorities.
