Digitalgipfel: KI als Kriegswaffe? Europa drohen Exportkontrollen

Arthur Mensch, co-founder of Mistral AI, warned during the Digitalgipfel in Berlin that US and Chinese AI models may be classified as weapons and subject to export controls. This could prevent the next generation of AI models from being supplied to Europe due to export restrictions. The classification stems from the "dual-use" potential of advanced AI technology, which can be used in both civilian and military applications.

The US has already enacted stringent export controls on high-performance AI chips and advanced semiconductor manufacturing techniques, which limit or prohibit the export to China. Mensch emphasized the danger of entrusting the development of large language models (LLMs) solely to the US and China, urging Europe to be able to produce its own "reasoning" AI models.

He stressed that Europe must possess its own model-making facility to avoid dependence on foreign suppliers whose power is continuously growing. Mensch compared AI development to the establishment of civilian nuclear energy, highlighting that AI is not just a software issue but a massive industrial undertaking. Tanja Rückert, head of digital at Robert Bosch GmbH, pointed out that Europe should leverage its strengths in industrial applications, particularly by creating "World Models" trained on real industrial data from factories and vehicles.

She emphasized that such data is not available online and would provide a significant competitive advantage for secure and reliable industrial AI. Robin Rombach, co-founder and CEO of the German AI startup Black Forest Labs, highlighted the importance of visual world models that learn intuitive physics. These models are crucial for robotics and could be executed locally on devices, without relying on cloud services.