September was the month the big AI labs stopped treating Australia as a sales territory and started treating it as a place to build, lobby and answer questions. Anthropic put pen to paper on local infrastructure, Canberra began sketching how the country might capture the gains, and the fight over copyright moved from op-eds to formal submissions. It ended with an apology from OpenAI and a Senate inquiry that wanted the CEOs in the room. Here are the stories that mattered.
Anthropic puts down roots
The biggest signal of the month came on 15 September, when Anthropic signed its first Australian data centre agreement (Reuters). For a country that has mostly consumed frontier AI from offshore servers, a lab of Anthropic’s standing committing to local capacity is a meaningful step.
A week later the company went public with the argument behind it, making its case for Australian AI investment in the AFR.
Canberra starts drawing the map
Government thinking moved forward too. On 18 September Canberra revealed its early thinking on how to capture the gains from AI (InnovationAus), and within days experts were urging an industrial focus for the national AI agenda.
Internationally, Australia joined a 22-nation push for global AI rules. ASPI’s Strategist added two sharp pieces to the debate: one arguing that Australia’s best bet for frontier AI access is a “coalition of the dependent” with other middle powers, and another warning that military AI is the next interoperability challenge as Australia spends on its own capability and on its Indo-Pacific partners’.
The copyright fight goes formal
On 22 September OpenAI and Anthropic formalised their push for Australian copyright reform (Capital Brief), calling for Australia to relax its ban on training AI models on copyrighted material.
The creative sector pushed back. ScreenHub argued that an opt-out system for AI companies doesn’t gel with Australia’s copyright laws, because it forces creators to take action just to protect rights they already hold under the Copyright Act. That argument would come back in October.
The data centre build keeps rolling
Below the headlines, the physical build-out carried on. Kalkine looked at whether Centuria Capital’s ResetData bet has changed the fund manager’s investment story, and the sector now has enough momentum for an inaugural Data Centre Build Expo, with Westkon Precast signing on as platinum sponsor.
The harder headlines
The month closed on accountability. On 26 September the CEOs of OpenAI and Anthropic were asked to appear before an Australian Senate inquiry. Two days later Bloomberg reported that Sam Altman would not attend.
On 30 September OpenAI apologised to Australia for AI breaches, and the AFR reported the company would turn Australia into a test lab in the fight against rogue agents. Information Age noted it unveiled autonomous agents just days after that apology.
Government security was in the frame too. Pinsent Masons examined how an AI breach of an Australian government website raises questions of liability, while Startup Daily reported the government is looking to harden its legacy technology against AI-driven threats.
What it adds up to
Put together, September showed Australia finally getting a seat at the table: a frontier lab building here, a government with an emerging plan, and lawmakers willing to ask the hard questions in public. The open items for the months ahead are clear enough. How the copyright fight lands, whether the infrastructure money turns into capacity, and whether the labs show up when the Senate calls. For what happened next, see our October roundup.
Comments 1