Hi! I created an account to respond specifically to this. I think it’s the wrong way of thinking about Middle Powers.
Let’s just focus on Australia for the moment. The United States is Australia’s strongest security partner. Australia is the junior partner in this equation, as it is in almost every other part of the relationship. If the Prime Minister decided tomorrow that AI was existentially going to “threat[en] our sovereignty…and even our lives”, declaring a national emergency would require an intense set of trade offs not really touched on or considered by the OP, and, critically, would achieve what exactly, if access wasn’t granted? How would the PM insulate Australia from a US response? Wake up moments seem the wrong way to approach this: July's Hugging Face breach was discussed in every G7 leaders' office, and what followed was a California subpoena rather than firmer national action.
Regardless, if the PM did this, he would immediately have to justify to both his population AND political class why these ‘costly signals’ are worth it, and more critically, why the immediate US or others response (not likely to be good if you’re trying to strong-arm access to a technology it sees as national security critical!) is worth the pain. Perhaps the Aussie PM could turn to his fellow ‘Middle Powers’ leaders for solidarity - but they have their own national interests and autonomous ideas driving them. In fact, some may see a market gap open. Others have significant exposure on other critical factors that US could lean on. How has the history to date of so-called Middle Powers doing this with other things—e.g., nuclear weapons - worked out? Compare these with agreements led by US, which tend to have a higher success rate.
It is incredibly, incredibly, incredibly difficult to agree on rhetoric, let alone ‘costly signals’ behaviour. IMO, this approach places far too much emphasis on the power of a Prime Minister/President to both control their own country AND navigate the US fallout, as well as the potential leverage cutting off your market presents.
The idea is that these leaders must believe RSI/Loss of Control/superintelligence are real, short-term threats. But consider being in the leader’s office: you are briefed by multiple experts, some inside and some external, who give you a series of views on how threatening this may or may not be. Every action has a reaction. Are you going to pull the trigger? What will the media say? The politicians? Your economy writ-large? How will the population weather the economic downturn?
Let’s say our leader decides to pull the trigger regardless. Would it be worth it for only getting an AISI/third party to investigate? And summits take months to prepare, so I find that confusing. Let us say they do, and…nothing happens on the US side, or among peer Middle Powers countries. How do you explain that to your electorate?
Politics is the messiest profession there is, made even messier when it involves geopolitics, especially for a non-tier one country. Leaders of so-called Middle Powers do not agree with each other on technical things even if they agree on broad strokes. Look at Ukraine: many European commitments have fallen through despite a LITERAL WAR on their continent and many of them calling it existentially concerning. Another example: in 2018, Australia, New Zealand and their Pacific partners (nations they have spent significant capital and aid supporting/culturally bonding with) endorsed the Boe Declaration, calling climate change the region’s “single greatest threat”. A year later Aus made sure the communique dropped calls for a rapid phaseout of coal and it wasn’t mentioned.
I suspect there are other options for what Australia or other Middle Powers can do that don't involve this specific framework.