Cohere chief urges G7 to keep control of national AI systems

Cohere CEO Aidan Gomez is pressing G7 leaders to treat artificial intelligence as a matter of national control, not only commercial competition. In a Fortune commentary published June 17, Gomez argued that governments and key industries could become exposed if access to major AI systems rests with a small group of outside providers.

Gomez pointed to what he described as a sudden restriction of access to Anthropic’s models as evidence of that risk. He said the episode showed how countries can lose access to important digital tools through a corporate decision or geopolitical shift.

The Cohere chief said he joined global leaders and executives at the G7 summit this week to discuss AI policy. His message, according to the commentary, was that governments should not “rent” core AI capacity from centralized vendors if doing so means giving up control over access, privacy and security rules.

Gomez compared AI dependence with earlier concerns about reliance on single energy suppliers or limited sources of critical minerals. He said democracies understood those risks in industrial supply chains and now face a similar problem with digital intelligence.

According to Gomez, autocratic states are promoting subsidized, state-influenced AI models that can extend centralized influence abroad. He said democratic countries should not answer that with dependence on another small set of closed systems controlled by private companies.

Gomez argued that a more resilient approach would give countries multiple providers while allowing them to preserve their own laws, languages and values. He said Cohere’s business reflects that view because it offers models that enterprises and governments can hold and run inside their own secure environments.

He also cited a technology alliance between Canada and Germany as a model for countries seeking advanced AI while keeping national decision-making authority. Gomez did not frame the issue as limited to Cohere, saying the wider question is whether countries can build real digital sovereignty rather than rely on marketing claims.

Gomez set out three tests that, in his view, governments and organizations should use to judge whether they have meaningful autonomy over AI systems.

• First, he said users need control over model quality, governance and updates, including the ability to tune systems for local requirements and decide when changes happen.
• Second, he said data control requires more than storing information in a local data center. He argued that the infrastructure should be located in the country and that outside access to the software or hardware behind critical systems should be restricted.
• Third, he said customers need operational independence, including the ability to run the same system in different clouds or on-premise environments if a provider’s outside services are shut down.

Gomez said digital sovereignty comes down to deciding who can view data, alter systems and cut off access. He urged the G7 to back advanced research, open environments and more flexible supply chains as it sets standards for AI resilience.

Fortune identified the piece as commentary and said the views were Gomez’s, not necessarily those of the publication.

This story draws on original reporting from Fortune.