Europe urged to reduce reliance on US AI infrastructure

Europe faces growing pressure to rethink its AI strategy after Anthropic introduced Mythos, a frontier model described as capable of hacking sensitive software worldwide. The US administration’s decision to withhold access to Mythos from non-US nationals exposed Europe’s dependence on American AI systems and showed that an estimated 88% of the world’s population could lose access to US frontier models overnight.

The central concern is that Europe is falling between two AI powers, with China projected to gain similar capabilities within a year. Project Glasswing, created to research and contain the risks linked to Mythos, includes no European actors, underscoring the gap between Europe’s regulatory ambitions and its technical capacity.

Proposed responses include mobilising major European companies for large private investment, completing the EU capital market Union and single digital market, and pursuing shared infrastructure such as European gigafactories and space constellations. Brussels is also urged to pool compute capacity with international partners and use multilateral cooperation as a counterweight to digital imperialism.

Europe is expected to remain dependent on US compute and chip infrastructure in the medium term, while using limited leverage such as ASML’s chip printing technology to negotiate a more balanced relationship. Public trust is framed as essential, with energy efficiency, water use and fairer distribution of AI gains positioned as prerequisites for rapid investment.