G7 leaders have started discussions on a proposed framework that would let a select group of allied nations access advanced US artificial intelligence models — a move that follows sweeping restrictions on AI exports and could redraw global tech alliances. The framework, described as a 'trusted partners' system, is still in early talks but could reshape how the United States shares its most powerful AI tools with close allies while keeping them out of the hands of rivals.
What the framework would do
The idea is straightforward: create a tiered access system where countries deemed trustworthy — likely current G7 members plus a few others — get preferential or even exclusive access to US-made AI models. Those outside the circle could face continued export controls or outright bans. The proposal comes as the Biden administration has clamped down on AI chip and model exports to China and other countries, citing national security risks. A 'trusted partners' label would effectively formalize and expand that approach, turning ad hoc restrictions into a structured alliance-based regime.
Why the G7 is discussing it now
The timing is no accident. US restrictions on AI model exports, announced in October 2023 and tightened since, have created friction even among allies who worry about being cut off from cutting-edge technology. Japan and the UK, for example, have both pushed for clearer rules on what they can and cannot access. By putting a 'trusted partners' proposal on the G7 table, US officials are signaling they want to keep close friends in the loop — but on Washington's terms. The discussions took place during the G7 leaders' meeting in Italy, where AI governance was a top agenda item. No final decision has been made, but the framework is expected to be debated further at the next G7 summit.
Global tech alliances in play
If adopted, the framework could shift geopolitical dynamics in a few ways. It would deepen the tech divide between the US-led bloc and countries like China, which is already building its own AI ecosystem. It could also force middle powers — India, Brazil, South Korea — to choose sides or risk being locked out of the most advanced US models. International AI collaboration, already strained by export controls, would become even more fragmented. Some G7 members have privately raised concerns that the system might alienate key partners or create a two-speed AI world. For now, those worries are being aired behind closed doors.
The biggest unresolved question is how the US will define 'trusted'. Will it be a short list of Five Eyes nations, or something broader? And what happens to countries that currently have access to US models but don't make the cut? The G7 leaders didn't set a deadline, but the framework is likely to resurface at the next summit — and possibly in bilateral talks before then.
