Cohere CEO on G7 leaders’ choice: sovereign AI or digital serfdom

The sudden restriction of access to Anthropic’s models exposes a reality many global leaders have chosen to ignore. We are in the early stages of an AI industrial revolution, yet nations remain dangerously dependent on a handful of big tech companies. Digital sovereignty goes far beyond market competition. It is fundamentally about who controls the technology that will shape our economic security and national autonomy for decades. This week, I joined global leaders and executives at the G7 summit to discuss the best way forward for artificial intelligence. My goal was to leave them with one clear message: The world can no longer afford the strategic risk of renting its future from centralized providers. Renting artificial intelligence means surrendering operational control. You give up your data privacy, your security protocols, and your basic access, leaving yourself entirely at the mercy of a third party. In the industrial era, democracies understood the immense danger of relying on a single source for energy or a narrow geographic choke point for critical minerals. Today, we are on the verge of repeating that exact mistake with digital intelligence. Autocratic nations are already leveraging heavily subsidized, state-influenced models to export an architecture of centralized influence. The alternative for the rest of the world cannot simply be a different form of dependency on a few centralized black boxes that decide when to provide or revoke access. A monopoly of intelligence is inherently brittle. When a handful of centralized entities control the primary models of the world, they control the parameters of global commerce, security, and thought. A single corporate policy change or geopolitical shift can instantly sever access to vital systems. We have to cultivate a competitive ecosystem where choice and control is guaranteed, allowing nations to rely on diverse providers while maintaining their distinct values, languages, and laws. This philosophy of distribut