Australia doesn't need sovereign AI.

American writer M. Stanton Evans once remarked that "bipartisanship" is when the country's "two parties get together to do something that's evil and stupid".

So it should be a little worrying that on the subject of AI sovereignty in Australia – i.e., whether the country needs its own model – major players of both the left and right factions are in broad agreement.

First up was senior Liberal Party member Andrew Hastie, who said "we are going to be a supplicant state, not a sovereign state":

"People think it's making memes on ChatGPT, (but) we're talking about the global order being changed and Australia falling behind. I want us to stay in front."

Then only a couple of days later former Labor Prime Minister Kevin Rudd said:

"More must come if we are to develop the sovereign AI capabilities set in this country that we will need for the future. We were once called the lucky country. We have now become often the complacent country... If we in Australia miss the AI revolution by not building our own sovereign companies, capabilities and competencies, the brutal reality is that Australia will no longer be counted in the first rank of nations."

I'm not sure if Rudd is aware of the full 1964 Donald Horne quote he paraphrases, because Horne was using "lucky country" as a pejorative:

"Australia is a lucky country run mainly by second rate people who share its luck. It lives on other people's ideas, and, although its ordinary people are adaptable, most of its leaders (in all fields) so lack curiosity about the events that surround them that they are often taken by surprise."

The good news is that Australia's sovereign status is no more dependent upon the second rate people running the country today than it was in 1964, even with the advent of AI. Australia needs its "own" frontier large language model about as much as it needs its "own" domestic car industry. Hastie in particular might disagree about the cars, but only because his economic thinking is stuck in the 1700s.

There are three main reasons why Australia will be just fine as a sovereign without its own AI. First, the AI space is fiercely competitive, and many of the leading models are open-weight. That could change, but right now it's relatively straightforward to host a top-tier AI on hardware you control that's fully sovereign. Don't like the leading Chinese models? Use Meta's Muse. The fact is the weights are public and no foreign government can cut off access. In that sense, Australia already has sovereign AI.

Second, these things are expensive: training costs are currently between US$1.5 billion and US$3 billion for a GPT-5 equivalent. That's assuming you already have a team and the hardware in place to actually execute a large-scale training run. And for what? To produce a model that will be at the frontier for maybe 7 months, with limited commercial opportunities because it would be competing against entrenched players with billion-dollar distribution networks, and freely available open-weight alternatives that already match or exceed anything an Australian government-funded lab is likely to produce.

Three, there's so much more to a sovereign nation than producing "its own" products. Contra Rudd, a lack of a home-grown AI model won't be what drops Australia out of the "first rank of nations". In fact, the opposite is likely to be true: every dollar Canberra diverts towards training a model that Meta will give away for free is a dollar not spent on the boring infrastructure that actually makes an economy work.

To remain in "the first rank of nations", Australia's leaders should avoid listening to Hastie and Rudd and instead make it easier for people to get on with the unglamorous work of adoption and diffusion, which is what will actually grow the economy.