Meta will not launch its new AI model due to EU data protection reasons

The strict EU data protection regulation have long posed a challenge between US tech giants and the European Union. Recently, Meta, the owner-operator of Facebook and Instagram, decided to withhold its next multimodal artificial intelligence model, Llama, from much of Europe, citing the stringent regulations. This model is designed to reason through video, voice, images and text, and will be integrated into a range of products, including smartphones and the Ray-Ban smart glasses developed by Meta.

This decision also means that European companies will not have access to the multimodal models even though they are being released under an open license. It may also prevent companies outside of the EU from offering products and services in Europe that utilize the new multimodal models. However, Meta has announced to release a larger, text-only version of its Llama 3 model, which will be available to customers and companies in the EU. The decision to withhold the Llama model is not due to the recently adopted European AI Act, but rather the General Data Protection Regulation (GDPR). Given the strict rules of GDPR, the company cannot train its models using data from users of its social networking platforms. In May 2024, Meta announced plans to launch a program in June to train its AI using publicly posted Facebook and Instagram content, but the program was halted in the same month when several regional data protection authorities began raising concerns.

Meta believes that training on European data is essential to ensure its products accurately reflect the region's terminology and culture. The company also noted that competitors such as Google and OpenAI are already training on European data. Interestingly, while the United Kingdom has a law nearly identical to GDPR, Meta has not faced the same level of regulatory uncertainty there and plans to launch its new Llama model for UK users. The company attributes this to the European regulators taking longer to interpret existing laws compared to their counterparts in other regions.

According to the European Centre for Digital Rights, European regulations would in fact allow Meta to introduce its AI models, but the company would need to obtain explicit consent from users first. However, Meta's current model assumes consent by default and requires users to opt-out if they do not wish to participate. This approach is problematic, as many users may not even be aware of how their data is being used, particularly since Meta has not provided clear notifications beyond a passing mention in its terms and conditions. Nonetheless, Meta has not abandoned its plans to introduce its AI model in the EU and is working on a solution that will be mutually acceptable to regulators and users alike. However, the impact on users remains unclear.