On a Friday afternoon in June, the US demonstrated just how little control Australia has over access to the AI models we’re increasingly reliant on. Anthropic received an order to suspend access to its most powerful models, Fable 5 and Mythos 5, for all foreign nationals.  Related Article Block Placeholder Article ID: 338862 Why the US just blocked Australians from using Anthropic’s most powerful AI Tegan Jones Within hours, those models disappeared for users outside US jurisdiction – only days after Mythos was first made available. And it wasn’t long before it happened again. OpenAI announced GPT‑5.6, a new model family led by its flagship Sol, designed to handle more complex coding and cybersecurity tasks.  Smarter business news. Straight to your inbox. For startup founders, small businesses and leaders. Build sharper instincts and better strategy by learning from Australia’s smartest business minds. Sign up for free. * indicates required Email Address * By continuing, you agree to our Terms & Conditions and Privacy Policy. But instead of a broad launch, OpenAI said access would start with about 20 “trusted partners”, pre‑approved by the Trump administration, after the White House asked it to limit the release.  OpenAI said in its announcement it doesn’t want this sort of government access process to become the default but is accepting it as a short-term step. At the same time, Anthropic has been warning Washington about what others were doing with its models.  In a letter to US senators, it accused operators linked to Alibaba of running the “largest known distillation attack” on Claude by using tens of thousands of fake accounts and millions of queries to train their own systems.  Anthropic framed this as “illicit” capability theft that turned US AI investment into a subsidy for a geopolitical rival and asked Congress to crack down. In both the Anthropic and OpenAI cases, national security decisions in Washington are now shaping which tools Australian organisations can use and when.  What has been made abundantly clear is that all the talk about AI being borderless and the great equaliser is bogus. Access and control of these models come firmly under US jurisdiction, no matter where in the world they’re being used. They’re also not mere neutral cloud tools. Frontier models are being treated as strategic assets whose release, access, and even imitation are questions of state interest. For Simon Kriss, CEO of Sovereign Australia AI, that should end any illusions about how much control Australia really has.  “I see this as Australia’s wake-up call. So long as we rent our future from the US or China, we are at their whim,” he told SmartCompany.  He also points out the practical cost. Governments and businesses may need “modern-day switchboards” that can detect when a model disappears and instantly flip to another. Related Article Block Placeholder Article ID: 338762 Neural Notes: OpenAI’s IPO is a reminder your AI strategy has a landlord Tegan Jones Kriss says this is “not cheap, easy or quick” and may change the economics of rolling out agentic AI processes at scale.  Across the G7, these US orders have already been read as a warning shot. Governments in Europe and Canada have pointed to the Anthropic incident as justification for accelerating their own sovereign AI and compute plans. It’s worth noting that as this column was being finalised, the US Commerce Department lifted the export controls on Anthropic, less than three weeks after imposing them. This follows a partial rollback over the weekend where, similar to OpenAI, the government allowed Mythos 5 to be reinstated for a handful of “trusted” US organisations. Now Anthropic will start restoring access more broadly, after agreeing to tighter security protocols and early notification of any “malicious activity”. However, the fact that Anthropic’s export controls were lifted within weeks doesn’t make the situation less serious. Both shutdowns and restorations of frontier AI access happen on someone else’s timetable. AI sovereignty versus agency In Canberra, the term that keeps cropping up is “sovereign AI”.  There’s been an uptick in ministers and shadow ministers alike working the phrase into speeches about national capability and the digital economy in recent years.  Just last week, Industry Minister Tim Ayres warned Australia cannot afford to be “a cork bobbing on the ocean of other people’s technology” and should secure as much of the technology stack as possible. Opposition figures like Andrew Hastie have started framing AI as a strategic contest reminiscent of the Cold War arms race. Hastie warned under-investment risks constraining Australia’s sovereignty and independence. Researchers are trying to give that rhetoric some tangibility.  The Automated Decision-Making and Society centre breaks sovereignty down into four elements: data, compute, model and policy. Across those metrics, Australia has seen patches of progress. There are on‑shore cloud and hosting providers, government-focused platforms, and a growing pipeline of AI-friendly data centres.  What the Anthropic shutdown exposed is that at the frontier, Australia still lacks model and policy sovereignty when it comes to the LLMs that are being sold to us as state of the art. Decisions about access sit with another government. The Tech Policy Design Institute (TPDi) refers to this as ‘AI agency’. This is a country’s ability to control outcomes, protect its interests and capture value from AI across infrastructure, data, models, skills, innovation and governance.  The TPDi’s recent report on the issue assesses 103 capability areas across six layers, finds that Australia’s agency is constrained by gaps in frontier models and domestic compute, even as it scores relatively better on skills, innovation and adoption. But Kriss is sceptical of attempts to soften the language. “Many people in Australia are using the word ‘agency’ to get around an inconvenient conversation on true sovereignty,” he said. “How can you have any form of agency if someone in another country can just turn off the tap?” Inside Sovereign Australia AI, he talks about five levels of sovereignty – infrastructure, models, agentic tools, applications and governance. But the underlying idea is relatively straightforward – GPUs and servers owned and operated here, models built and run here, and rules set here. Related Article Block Placeholder Article ID: 322958 Neural Notes: Sovereign Australia AI believes LLMs can be ethical Tegan Jones “The only answer is our own AI on our own servers that is owned and operated by Australian entities only,” Kriss said.  Some builders, however, have lost patience with the glossary of terms. Maincode’s view on dependence Maincode, which has created the Matilda model as well as a Melbourne-based ‘AI factory’, stopped using the term ‘sovereign AI’ last year. “We ditched the term because it doesn’t mean anything. It’s not the language of builders. You can’t test it; it holds no value, no meaning,” Maincode CEO Dave Lemphers told SmartCompany.  Instead, Maincode talks about “Australian-made AI” that has an end‑to‑end stack built and run locally. Lemphers says the Anthropic order did not materially affect Maincode because the company is not relying on US-hosted frontier models in the first place. Instead, it runs on its own models that he claims outperform Claude. “We use our own models and consume our own tokens, from chat through Matilda to our Matilda-code coding agent,” Lemphers said “I’ll admit it does concern me, though, that there are still Australian organisations, both government and enterprise, comfortable with a foreign political entity like the White House dictating what AI they get to use. “For us at Maincode, it’s liberating to not have to worry about access to state-of-the-art AI being dictated by someone else.” The company has detailed publicly how it wrote its own inference engine in Rust for AMD Instinct GPUs, bypassing the standard vendor stack. It also argues owning an on‑premise “AI token factory” makes tokens far cheaper than buying them through public APIs. “The reality is, we’re doing it, and doing it well. People talk about it in terms of ‘what if’. We’re so past that. We’re actually doing it, and I think the conversation is lagging behind the reality.” Lemphers doesn’t mince words when it comes to Australia’s reliance on offshore models. “The idea that using foreign-controlled AI is completely fine goes far beyond commercial considerations. It is the single biggest standalone risk to Australia’s national security posture and our economic ability to compete globally,” he said. “I just don’t think those in power actually get it, to be honest.” What this means for Australian businesses This might sound like a story for policymakers. It’s not. This sits firmly in the territory of how Australian businesses buy and use AI. If a company has built workflows around a specific US‑hosted model, it has assumed that model will be there on roughly the same terms tomorrow. The Anthropic shutdown showed that assumption can be overturned overnight. If an AI vendor is selling itself as Australian or “on‑soil” but cannot say what happens when US or other export rules change, that’s a red flag. Related Article Block Placeholder Article ID: 339545 We grew up fearing piracy laws, now AI is rewriting them in Australia Tegan Jones Rather than vague political language around AI that “serves Australians,” we need to more clearly define what sovereign or even ‘Australian-made” actually means when it comes to infrastructure and models. Kriss worries that “companies are calling themselves sovereign and touting being sovereign when in fact they are not”, pointing to structures where an ostensibly Australian entity is owned by a US company and still covered by The US CLOUD Act. “What does it take for a data centre to be sovereign? Or an app developer? Or a model? Is it just that the majority of shareholders are Australian? Or is it that the Board is based here? Or is that profits and taxes stay here? This is not an easy task but one worth doing,” Kriss said. Frameworks such as the TPDi’s AI Agency tool and ADMS’s sovereignty checklist do give those conversations something more tangible to work with. But we need to get honest about where we are willing to depend on foreign AI, and build enough backup options so Australian organisations aren’t blindsided when another government revokes access. Neural Notes is a weekly AI column that looks at how AI is influencing Australia. Stay in the know Never miss a story: sign up to SmartCompany’s free daily newsletter and find our best stories on LinkedIn.