Макрон и Моди предупредили G7: США могут отключить AI в любой момент

The leaders of France and India publicly warned at the G7 summit: dependence on American AI platforms is not an abstract geopolitical risk, but a real vulnerability that can be activated by a single decision from Washington. The shutdown of Anthropic services, which occurred during the same period, only confirmed their point.

What was said at G7

Emmanuel Macron and Narendra Modi raised the same question in the corridors of the summit: no country relying on American AI tools can be confident in the continuity of their operation. One political decision in Washington, a shift in regulatory course, or a technical failure would be enough — and critical infrastructure of an entire country would become unavailable. This is not a theoretical scenario. The Biden and Trump administrations have already used export controls to restrict access to NVIDIA chips — dozens of countries ended up on sanctions lists. Restrictions on AI services are the next logical step that no one in Washington officially rules out.

"We need access to the best AI.

But we cannot afford a situation where one call from Washington cuts off our access to tools our economy depends on."

Why Anthropic's blackout became a symbol

When Anthropic's services went down for several hours, it was an ordinary technical failure — not a political act. But in the context of the G7 discussion, it acquired political dimensions: this is what the world looks like where critical AI systems are managed by one corporation from one country. A corporate failure and a government shutdown are different things by nature.

But for the user on the outside, the result is identical: the service is unavailable, the data is unreachable, tasks are not performed. For governments that are integrating AI into state administration, this means paralysis of entire departments. It is this argument that is now being used by supporters of sovereign AI: don't wait for political will — if it's already happening technically by accident.

Sovereign AI: from words to money

The term "sovereign AI" appeared on the agenda after 2022, but for a long time remained at the level of declarations. Now budgets have started moving:

• France has invested over €2 billion in national AI laboratories and the development of Mistral as a European alternative to GPT
• India has launched the IndiaAI program with a budget of $1.25 billion and is building Bhashini — a platform for working with local languages
• The EU adopted the AI Act, partly aimed at reducing dependence on American providers
• Saudi Arabia is building an AI cluster with 500 MW capacity jointly with NVIDIA — with a clause on national control over data
• China is already de facto operating in an isolated AI space with its own foundational models

The paradox is that most of these projects still depend on American chips, frameworks, and basic architectures. Complete independence is a horizon goal, not one of the next quarters. But even partial diversification reduces risks.

What this means

AI sovereignty is becoming as much a part of the foreign policy agenda as energy independence following the 2022 crisis. Governments are moving from words to concrete budgets. For American AI companies, this is a new reality: they will have to offer guarantees of availability, local data storage, and independent management — otherwise states will choose less technologically advanced, but more controlled alternatives.