Particle.news

U.S. Export Controls Cut Europe Off From Anthropic’s Top Models as EU Moves to Build Its Own Frontier AI

The step has prompted a rush to fund European compute, chip and regulatory capacity to prevent future U.S. restrictions from severing access to leading AI technology.

Overview

  • Anthropic has disabled access to its Mythos 5 and Fable 5 models for non‑U.S. users after U.S. authorities ordered limits on foreign access, a decision reported Friday that left many European researchers and companies without the newest tools.
  • U.S. officials cited security worries, including demonstrations by outside researchers that Fable’s safeguards could be bypassed and concerns about some partners’ links to China, which together prompted the government to press for export controls.
  • Brussels announced the winner of its Frontier AI Grand Challenge, the consortium ‘Europa’, which will receive EuroHPC resources to develop an open 400+ billion‑parameter model, while Germany and France agreed to set up a joint DFKIInria AI centre to accelerate sovereign capacity.
  • Analysts and officials warn Europe holds under 5% of global high‑performance AI compute compared with roughly 75% in the U.S. and about 15% in China, a gap that leaders say requires faster permits, energy and chip investment to fix.
  • Public trust is falling even as usage rises, a dynamic that could slow adoption of new European systems and makes rapid, visible steps on safety, transparency and local procurement crucial to keep businesses and governments as customers.