The U.S. government told Anthropic, a major AI company, to shut off access to its best AI models for all users outside the country on June 13. Then, just days later, leaders from the G7 group of nations met in France to discuss a new system: which allied countries should be allowed to use these powerful AI tools. Reuters reported on the G7 discussions on June 16.

Anthropc quickly shut down its Fable 5 and Mythos 5 models for everyone outside the U.S., according to a June 13 Reuters report. PBS NewsHour described this as a White House security directive.

Now the G7 is working on what they are calling a "trusted partner" system. Instead of allowing everyone or no one, this approach would let the U.S. choose which countries get access based on whether they are reliable allies. Think of it like a guest list — some countries make the cut, others do not. The specific rules for how this will work have not been released publicly.

Why does this matter? Powerful AI models have become something governments treat like military or strategic assets. The suspension and the G7 proposal show that the U.S. now sees advanced AI as a national security tool, not just a technology companies can sell freely. Other countries are locked out unless the U.S. decides to let them in.

The European Union adds a wrinkle to this plan. Some EU member countries will likely be trusted partners, but others may not be. The EU does not speak with one voice in these negotiations, so individual countries could end up treated differently. That could create tension within the alliance itself.

Anthropc's quick obedience to the government order is worth noting. The company is not a traditional defense contractor. Yet it complied within days, which suggests either the White House had planned this in advance or the company felt it had no real choice. Either way, other AI companies are now watching to see what similar orders they might receive.

This approach borrows from how the U.S. has controlled semiconductor chip exports for years. Chip controls started broad, then became more selective over time. AI controls are following a similar path, but the rules are still being written. A summit discussion is a political signal, but it does not mean a real system is in place yet.

The bottom line: advanced AI models are now treated by governments as strategic tools, just like nuclear technology or advanced weapons. Countries that the U.S. trusts will get access. Others will not. For companies locked out right now, the waiting period could be long.