Five Eyes warn AI cyberattacks are imminent as US freezes Anthropic Fable access

The Five Eyes warning and the Anthropic export crackdown are a watershed moment in how states treat frontier models: less like consumer software, more like dual‑use munitions. Intelligence agencies are effectively saying that the next big jump in offensive cyber capability will be driven not by human hackers but by agentic AI systems that can discover, chain and exploit vulnerabilities at machine speed. The fact that the U.S. was willing to unilaterally cut allies off from Anthropic’s top models underscores how seriously Washington now views that threat surface.

From an industry perspective, the move shatters the assumption that commercial frontier access will be globally homogeneous. If nationality and alliance status determine who can touch the most capable models, we’re looking at a fragmented AI order where Five Eyes, EU, Chinese and other blocs increasingly rely on their own sovereign stacks. That pushes companies and countries without favored access toward either domestically built models or open‑weight systems they can fully control.

For the AGI race, this is a double‑edged sword. On the one hand, restrictive export controls and chilling effects on cross‑border research could slow diffusion of the most capable systems. On the other, they create powerful incentives—especially outside the U.S.—to pour capital into alternative stacks, potentially accelerating a multipolar frontier landscape rather than a single, U.S.-centric trajectory.