AI & EU: time to empower innovators

Inspired by an open letter "Europe needs regulatory certainty on AI"

Countries can be leading in culture and sciences before they fall back for decades or even centuries. The Roman Empire is one of the many examples of human history that could see the rise and fall of nations. More recently, Japan’s nearly two-and-a-half centuries of isolationism de facto froze progress in the far eastern archipelago. 

Global competition and complex societal changes currently pressure Europe’s decision-makers to steer the continent toward competitiveness and innovation—at least on paper. 

One factor that might have a decisive impact on the EU's relevance and connectivity in the world is how permissive the EU’s rules for open-source artificial intelligence models will be.

Will European Data end up in AI Models?

The EU’s competitive edge in AI depends on its single market and consistent regulatory rules. Clear and stable regulations are necessary to encourage substantial investments in Generative AI and ensure the effective use of European data. 

However, recent regulatory uncertainty from data protection authorities has created significant uncertainty, particularly concerning data usage for training AI models. Companies and institutions are often not certain what data they can use to train these models. 

This uncertainty jeopardizes the development of open-source AI models and the creation of AI products and services that reflect European knowledge, culture, and languages. We need to ask the question if all future progress in data science and AI will be determined by byzantine data protection authorities or if we want to embrace a coherent framework that is actually operational and allows small and big, non-profit and for-profit players to advance technologies - and hence serve consumers.

Mr. Robot, Do you speak European?

Inconsistent regulatory decision-making is stifling its competitiveness and innovation. Companies and investors will consider refraining from European data and bluntly said, AI might just not speak European. Without a unified regulatory framework, the EU may miss out on two critical aspects of AI development: open models and multimodal models.

Imagine a future in which AI models lack input from European sources, literature, and users. From mundane applications such as cooking recipes to probably more important issues such as business risk assessment models, the European perspective might lack, making their application less impactful (and hence value-creating) than elsewhere in the world.

Open models are freely available for modification and use, fostering widespread benefits and opportunities. They enhance sovereignty by allowing organizations to adjust these models locally, avoiding the need to share sensitive data elsewhere. Multimodal models, which integrate text, images, and speech, represent a significant advancement, offering capabilities akin to having multiple senses compared to text-only models.

Frontier-level open models, whether text-based or multimodal, can significantly boost productivity, advance scientific research and contribute hundreds of billions of Euros to the European economy. They are already accelerating medical research, preserving languages, and providing essential tools to businesses and startups. If Europe fails to embrace these models, it risks missing out on technological progress enjoyed by the US, China, and India. 

Yet another train not to miss

Europe faces a critical decision that will shape its future. It can either uphold the principle of regulatory harmonization as outlined in frameworks like the GDPR, adapting its provisions to support AI innovation, or it can continue to hinder progress caused by fragmentation. If the latter persists, Europe risks falling behind while other regions advance and capitalize on emerging technologies. This could lead Europe to a period of AI isolationism.