US export controls cut Australians from Anthropic AI, delayed OpenAI release

A US government order blocked Australian access to Anthropic's top models in June. Days later, OpenAI delayed its GPT-5.6 launch, limiting rollout to 20 White House-approved partners. The moves highlight how little control ANZ has over AI infrastructure sales teams increasingly rely on for prospecting, research, and customer engagement.

US export controls cut Australians from Anthropic AI, delayed OpenAI release

What Happened

On a Friday in June, the US government ordered Anthropic to suspend foreign access to Fable 5 and Mythos 5, its most powerful models. Australian users lost access within hours, only days after Mythos launched.

Weeks later, OpenAI announced GPT-5.6 Sol, designed for complex coding and cybersecurity tasks. Instead of a broad launch, the White House asked OpenAI to limit access to roughly 20 pre-approved partners. OpenAI said it does not want this government approval process to become standard but is accepting the restriction.

Both moves stem from US export controls on AI chips and models, managed by the Bureau of Industry and Security (BIS). These controls determine which countries and entities can access advanced AI capabilities. Australia sits outside the approved tier, despite being a Five Eyes partner.

Why This Matters for Sales Teams

Most ANZ sales orgs now use AI tools for prospecting, email generation, call prep, and research. If you are running Anthropic or OpenAI tools in your stack, you just learned those can disappear on short notice.

The bigger issue: we have no visibility into when the next cut happens. No advance warning. No grandfather clause for existing customers. Your SDRs could lose access to their AI workflow mid-quarter.

This is not about one vendor. It is about dependency on US-controlled infrastructure with no ANZ fallback. Every go-to-market team building AI into their process needs a contingency plan.

What This Means for Sovereign AI

Australia keeps talking about sovereign AI capability. These export controls make the case stronger. If your sales stack depends on models that can be switched off by executive order, you have operational risk.

Global trade compliance software and AI compliance certification are becoming table stakes for any vendor selling into enterprise. Legal teams will start asking about jurisdiction, data residency, and export control exposure before approving new tools.

Worth noting: Neural Concept, a Swiss AI firm (not US, despite some headlines), just closed $100 million Series C led by Goldman Sachs. The funding round signals growing interest in non-US AI alternatives. Competitors include Ansys, Siemens, and Autodesk in the engineering AI space. ANZ presence is unclear, but the growth trajectory is worth watching.

The Bottom Line

If your sales team depends on US AI models, you just got reminded that access is conditional. Build redundancy into your stack or accept the risk that your tools could disappear mid-quarter.