Australians abruptly lost access to Anthropic’s most advanced AI models over the weekend after the US government ordered the company to suspend access for foreign nationals, citing national security concerns. The decision affects Anthropic’s Fable 5 and Mythos 5 models. The latter was only made available earlier this month to select Australian organisations through its Project Glasswing program. Related Article Block Placeholder Article ID: 338310 What does Anthropic’s Mythos AI actually do? Tegan Jones The reason for the suspension has been concerns regarding the models’ safety guardrails; however, there is arguably a bit more to it. Advanced AI models increasingly being treated as strategic technologies. This raises some important questions around who gets access to the world’s most powerful AI systems and under what conditions. And how easily they can be taken away. What are Fable and Mythos? Anthropic released Fable 5 and Mythos 5 as part of what it describes as a new generation of frontier AI systems. 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. Fable 5 is Anthropic’s flagship general-purpose model. It is designed to reason through complex problems, write software, conduct research, and perform sophisticated technical tasks with relatively little human guidance. Fable is a broad frontier model intended for everything from software engineering and scientific research to business analysis and planning. Anthropic has positioned it as one of its most capable AI systems to date. Mythos 5 is a specialised cybersecurity model built on similar underlying technology, but it’s focused on identifying software vulnerabilities and assisting security researchers in finding weaknesses before attackers do. Anthropic has repeatedly argued that Mythos is powerful enough to create new cybersecurity risks if released broadly. As SmartCompany previously reported, Anthropic claimed participating organisations had already identified more than 10,000 high- or critical-severity vulnerabilities across widely used software systems. Related Article Block Placeholder Article ID: 332633 Neural Notes: Australian AI startups may face the same defence dilemma as OpenAI, Anthropic Tegan Jones Why did the US government intervene in Anthropic? AI developers place guardrails around advanced models to prevent them from being used for harmful purposes such as cyberattacks, biological weapons research or other dangerous activities. According to a statement on the company’s website, Anthropic says the US government believes there may be a way to bypass or ‘jailbreak’ safeguards built into Fable 5 that are intended to stop the model being used to identify software vulnerabilities. The company disputes that assessment, saying there has been a misunderstanding and is working to restore access. According to Anthropic, officials have only provided what it described as verbal evidence of a narrow and limited method of bypassing restrictions rather than proof of a broader security failure. Even so, the government directed the company to suspend access for foreign nationals while the issue is investigated. The dispute comes only days after Anthropic CEO Dario Amodei called for stronger government oversight of frontier AI systems.  Anthropic argues this case does not meet that threshold, saying the government’s concerns relate to a narrow and disputed safeguard bypass rather than evidence of a broader safety failure. The company also said the order applies to foreign nationals broadly rather than distinguishing between close allies and geopolitical rivals. This is why Australian users were caught up alongside users from countries that have traditionally faced US technology restrictions. Some analysts have also argued that Five Eyes countries should receive different treatment from a blanket foreign-user ban. The shutdown also comes against the backdrop of a deteriorating relationship between Anthropic and parts of the Trump administration. Earlier this year, Anthropic publicly clashed with government officials over national security issues and reportedly refused to support certain military applications of its technology, including domestic surveillance and fully autonomous weapons systems. Related Article Block Placeholder Article ID: 334214 Anthropic leak lands hours after Australian government’s AI deal Tegan Jones The company was subsequently designated a supply-chain risk by the US government, a decision Anthropic is challenging. What it means for Australian businesses While relatively few Australian organisations had direct access to Mythos in particular, it does point to the very real platform risk for businesses relying on frontier models.  Access to advanced models may increasingly depend on export controls, nationality, and geopolitical considerations rather than just commercial decisions. Pricing and product feature changes are one thing. And we touched on that in Neural Notes last week regarding OpenAI’s IPO. But a government deciding whether certain capabilities should not be available outside their borders at all is quite another. And it will likely strengthen the increasing sovereign AI capability debate within our own borders. Stay in the know Never miss a story: sign up to SmartCompany’s free daily newsletter and find our best stories on LinkedIn.