G7 warned: Nations risk AI dependency on tech giants

What happened

Cohere's CEO addressed G7 leaders this week to argue that nations are becoming dangerously dependent on centralized AI providers. He warned that restricting access to models—citing the recent restriction of access to Anthropic's models—exposes a structural vulnerability, and called for countries to shift from "renting" AI from third parties to building their own sovereign systems.

Why it matters

When nations rely on centralized providers, they forfeit control over data privacy, security, and basic access to systems that will shape economic and national security for decades. A single corporate policy change or geopolitical shift can instantly sever access to vital systems. The CEO argues this mirrors historical mistakes democracies made by depending on single energy sources or geographic chokepoints for critical resources.

What to watch

The CEO outlined three practical tests for genuine sovereignty: control over model quality and updates; data infrastructure located within the nation with no third-party access to software or hardware; and the ability to run systems across different cloud or on-premise environments without dependence on a provider's services. He also referenced a tech alliance pioneered by Canada and Germany as a practical blueprint for nations seeking to access cutting-edge capability without trading away self-determination.