Washington just drew a hard border around the most advanced software on earth, and America's closest allies are scrambling.
If you've been tracking the global tech race, you know the big talking point used to be how the West would collectively beat China. Turns out, the house is divided. US President Donald Trump ordered Anthropic to pull the plug on foreign access to its latest models, Fable 5 and Mythos 5, over national security concerns. Just like that, developers, banks, and state agencies in London, Paris, and Tokyo were locked out of the best tools in the business.
This isn't about chips or supply chains anymore. It's about direct, sovereign control over the intelligence layer itself. At the G7 summit in Evian-les-Bains, France, the typical diplomatic pleasantries have been replaced by tense, backroom bartering. The US is turning code into a geopolitical weapon, and even its friends are getting hit by the shrapnel.
The Illusion of the Shared Tech Alliance
For years, the Group of Seven operated under a comfortable assumption. The US would build the frontier tech, and the broader democratic alliance would deploy it to defend against authoritarian cyber threats. That assumption died when Anthropic disabled foreign access.
The immediate fallout hits hardest in the UK and continental Europe. British Prime Minister Keir Starmer spent days trying to negotiate an exemption, only to get a firm rejection from the White House. Meanwhile, France is already reacting; the French government recently dropped Palantir for domestic alternatives specifically to break its dependency on American infrastructure.
The strategy behind the US block isn't subtle. White House officials and US Commerce Secretary Howard Lutnick are treating advanced artificial intelligence as a strictly national asset. Take Anthropic's Mythos 5, for example. It's a highly specialized model designed to scan computer code and identify critical vulnerabilities. In the wrong hands, it's a blueprint for catastrophic cyberattacks on banks, power grids, and water systems. Prior to the executive order, Anthropic had quietly granted access to organizations across 15 countries to secure their critical infrastructure. Now, those systems are in limbo.
The European Union is desperately trying to gain access to Mythos just to study its architectural risks, but Washington holds all the keys.
The Trusted Partner Trap
To quiet the uproar at the Evian-les-Bains summit, US representatives started floating a workaround: a "trusted partner" framework. The idea, which leaked via the Financial Times, would allow select foreign governments or specific corporations to regain access to advanced American models.
On paper, it sounds like a reasonable compromise. In reality, it's a geopolitical trap. Here is what's actually happening when you strip away the diplomatic spin:
- Absolute Dependency: A trusted partner status means your national cybersecurity apparatus operates entirely at the pleasure of the US executive branch. If a foreign government disagrees with Washington on trade, defense, or climate policy, that access can be revoked overnight.
- Economic Subsidization: Tech circles are starting to call out the financial imbalance. The current administration openly favors a model where Europeans pay premium licensing fees to American tech giants to use these systems, essentially funding the next generation of US models while starving domestic European innovation.
- Corporate Chokepoints: Silicon Valley executives are directly setting the terms of international security. CEOs like Sam Altman of OpenAI, Demis Hassabis of Google DeepMind, and Dario Amodei of Anthropic are sitting down for working lunches with G7 leaders to negotiate global regulatory guardrails. Private tech companies are effectively operating as sovereign entities in these talks.
Why Sovereign AI Isn't Optional Anymore
If you are a policy director or a tech executive outside the United States, the lesson from this G7 summit is crystal clear: relying on American APIs for your core infrastructure is a massive business risk.
We used to talk about "sovereign AI" as a vague, nationalistic ideal. Now it's a survival mechanism. When the US government can flip a switch and disable your tech stack because of an executive order, you don't actually own your infrastructure.
The immediate next step for international firms and allied governments isn't to beg for trusted partner status. It's to aggressively fund and deploy open-weights models and local compute infrastructure. You need to audit your current software stack, identify every single point of failure tied to American hosted models, and build out redundancy using independent infrastructure. Relying on Washington's ongoing goodwill is no longer a viable strategy for global business.
