Mistral chief urges European strategy for frontier AI
Arthur Mensch says Europe needs its own advanced models and computing capacity, warning that access to US and Chinese technology cannot be taken for granted.
Mistral chief executive Arthur Mensch called for a Europe-wide approach to developing advanced artificial intelligence on 2 November, arguing that industrial applications alone would leave the continent dependent on foreign technology. Speaking alongside the Digital Summit in Berlin, he said European security required investment in both models and computing capacity.
Mensch warned that possible US or Chinese export restrictions could limit European access to powerful AI models. He argued that Europe had sufficient talent but lacked the capital and computing resources needed to train sophisticated systems and develop companies capable of competing at scale.
His remarks coincided with the German government announcing an AI strategy for 2030. Alongside a focus on industrial uses, Germany plans to support the establishment of at least three frontier AI laboratories in Europe by that date. Mensch welcomed the ambition but said national action was insufficient.
In September, senior EU leaders advocated developing applications using foreign models rather than pursuing frontier models. The EU's 27 leaders are due to hold their first discussion devoted to AI at an informal Dublin meeting.
Mensch said manufacturing was Mistral's largest market, citing partnerships with Airbus and ASML. He nevertheless argued that industrial AI could not deliver technological independence while the underlying models and infrastructure remained under foreign control.
Competition on cost and performance remains a challenge. German search engine Ecosia stopped using Mistral's models; its chief executive said moving to Chinese alternatives halved costs and improved overall performance. Mensch did not directly address that experience, but rejected the suggestion that sovereignty required customers to sacrifice quality or competitiveness.
