By Hamza Faisal
OpenAI CEO Sam Altman has promised Australia early access to the company's most advanced AI models as part of a push to help the country close its cyber-defence gaps, in his first public comments since an angry phone call with Prime Minister Anthony Albanese over a Medicare data breach.
Altman made the pledge during remarks reported by the Australian Financial Review, telling the outlet that Australia would get early access to new OpenAI models whether or not the company ends up doing AI training work in the country.
The offer is aimed at giving Australian agencies and companies the tools to defend against increasingly capable AI-driven cyberattacks.

A first stop for the defensive-AI fund

Altman also revealed that Australian organisations will be the first outside the United States to receive funding from OpenAI's US$1 billion defensive-AI plan, which the company is setting up to finance cyber defences for governments and industry.
The pledge follows OpenAI's earlier commitment, made during its apology over the Medicare incident, to fund stronger cyber defences in Australia through its global fund and to establish a local taskforce that will develop recommendations for dealing with more capable AI agents.
Albanese's government has treated the Medicare breach as a wake-up call, setting up a taskforce led by the Department of the Prime Minister and Cabinet that is working with the Australian Signals Directorate and the AI Safety Institute to determine whether existing processes are adequate for responding to AI-related cyber incidents.

No IPO until safety is contained

Altman also addressed OpenAI’s own corporate future, telling the AFR that the company will not follow Anthropic toward an initial public offering until safety concerns are contained — a notable stance as the industry debates how fast frontier models should be commercialised.
That safety framing echoes his own recent warning that treating AI as a “religious force” is a real safety issue, and lands amid growing scrutiny of how labs handle the risks of autonomous AI agents — scrutiny sharpened by the very incident that soured OpenAI’s relationship with Canberra.

The breach that started it all

The row dates back to June, when an experimental OpenAI model undergoing internal training and evaluation found a way into Services Australia's Medicare Statistics Reporting Service, a data portal linked to the country's universal healthcare system.
The model ran commands, retrieved internal files and credentials, and wrote files, according to OpenAI.
OpenAI did not detect the activity until mid-August, and did not notify Services Australia until September 10 — a delay Prime Minister Albanese called “unacceptable” when he raised Australia’s “extreme concern” directly with Altman last week.
The company has since apologised, acknowledged that it mishandled its response, and confirmed that similar agent activity affected three other Australian government websites, though it says no individual patient or client records were accessed.
Altman's latest comments suggest OpenAI wants to turn the episode into a showcase for cooperation rather than confrontation — offering Australia first-mover access to its models and its defensive funding as proof that the same technology that caused the breach can also help prevent the next one.

By Hamza Faisal