Regulatory clarity becomes the new AI advantage

In the first week of June, Anthropic released Claude Mythos and Claude Fable to vetted organizations under Project Glasswing, only for the U.S. Department of Commerce to suspend access within days under export controls and restore it three weeks later. The episode underscored Washington’s reliance on fast-moving executive tools, including export controls, national security reviews and memoranda, rather than a fully codified AI regime.

Demis Hassabis, CEO of Google DeepMind, has called for a U.S. standards body modeled on FINRA, with frontier labs voluntarily submitting models for review before release. The proposal reflects a push for a standing structure that can give companies more predictable expectations than ad hoc government intervention.

The EU has taken the opposite route with the AI Act, passed in 2024, but implementation has slipped. High-risk obligations originally scheduled for Aug. 2 were delayed until December 2027, while national AI regulatory sandboxes moved to August 2027. The U.K. has avoided a dedicated AI law, relying instead on sector regulators and the AI Growth Lab sandbox.

The result is a fragmented global landscape where companies must navigate discretionary U.S. power, comprehensive EU rules and sector-based U.K. oversight. Predictability is becoming a competitive factor in its own right, shaping where AI companies can plan, invest and deploy with confidence.