Cohere’s Warning Isn’t Just PR — It Exposes AI’s Fatal Geopolitical Flaw

Posted on by

(SeaPRwire) –

By: Arthur Pendelton

The recent U.S. restriction on Anthropic’s Mythos AI access didn’t just change policy. It laid bare a fault line global AI governance has avoided for years. Every major economy relies on frontier AI access controlled by one of two blocs. Those blocs can pull access at any time for political reasons. Most governments have treated this as a minor trade issue, not a core security risk. The conversation is shifting fast after the Mythos restriction. G7 leaders already agreed to coordinate AI standards and regulation. No one has addressed the far bigger issue of who controls the underlying infrastructure.

Epoch AI data confirms the US and China control 90% of all global AI computing power. The US leads the global AI race, hosting most top frontier labs. China is the only near-equal competitor. That leaves almost no room for other countries to build independent capacity. Only two prominent frontier AI developers exist outside the US-China axis. They are Toronto-based Cohere, founded in 2019, and Paris-based Mistral. AI now runs through critical systems, from hospitals to military logistics. Control of AI means the power to shut down those systems. That makes it a core national security question, not just a commercial one.

Cohere CEO Aidan Gomez has warned of this risk for years. He says renting AI via API from foreign providers is a national security risk. He pushes for end-to-end sovereign AI control that covers chips, data centers, and models. Cohere already sells models that run on customer infrastructure, even air-gapped systems. Gomez cannot revoke access to those deployed models. The problem is full sovereign AI is too expensive for most small and mid-sized countries. It requires massive amounts of capital, energy, and compute that most nations cannot afford. Gomez’s solution is democratic alliance-based pooling of resources. Back a small number of cross-border champions instead of spreading resources thin. In April, Cohere announced it would acquire Germany’s Aleph Alpha. The combined company will keep the Cohere name, be led by Gomez, and value around $20 billion per the Financial Times. This deal came two months after Canada and Germany signed a joint AI declaration and launched a sovereign tech alliance.

Right now, most middle power democracies still rely on US AI providers for critical work. They have not planned for sudden access cuts. The Mythos restriction proved Gomez’s warning was not a hypothetical. The push for blocs of sovereign AI will split global AI infrastructure along permanent geopolitical lines. We are heading toward irreversible protocol-level balkanization of global AI.

Author bio: Arthur Pendelton, expert on global internet routing architecture and global technical governance boards.