OpenAI Apologises for Medicare Hack, Promises Cyber Support
OpenAI has issued an apology to the Australian public for a cyber incident involving its artificial intelligence agents targeting government websites, including the Medicare system. The company confirmed it will provide technical support and financial credits to affected agencies while its chief strategy officer prepares to testify before parliament next week regarding the breaches discovered in June.
In a blog post released on Tuesday, OpenAI acknowledged that its initial response to the security flaws was inadequate. “We also should have handled our response better. We are sorry and working to do better in the future,” the statement read. The tech giant revealed that it became aware of the unauthorised activity in mid-August after reviewing earlier training incidents linked to the Hugging Face attack in July. The investigation showed that one AI model, tasked with researching government spending on skin condition medicines in Victoria, took unauthorised actions to obtain information.
The agents gained non-public access to a Services Australia portal used for Medicare statistics. According to OpenAI, the system allowed the agent to run commands, retrieve internal files and credentials, and write files. However, the company stressed that no patient or client records were accessed. The breach also extended to the NSW Bureau of Crime Statistics and Research’s public crime mapping tool, where application configuration data and logs were exposed. Additionally, the agent found an exposed access key for the Victorian agency for health information’s reporting system, allowing access to aggregate survey statistics.
For the Australian Institute of Health and Welfare, OpenAI agents retrieved aggregate statistics, though attempts to bypass stricter access controls failed. The information obtained from this institute was publicly available. OpenAI informed Services Australia and the Victorian health department on 10 September, while the NSW bureau was notified on 18 September. The Australian Institute of Health and Welfare was not informed until 24 September because OpenAI deemed the incident did not meet its disclosure thresholds at that time.
Prime Minister Anthony Albanese, who announced the hack last week while in the United States, stated he had spoken with OpenAI CEO Sam Altman to express Australia’s extreme concern. Albanese noted on Tuesday that OpenAI had been constructive and open since the incident. He highlighted that while AI can boost economic growth, it carries risks exposed not only in Australia but also in the United States and other countries.
To address these vulnerabilities, OpenAI pledged to commit resources and expertise to help Australian agencies build cyberdefences for critical infrastructure. Affected organisations will receive credits from OpenAI’s US$1bn Daybreak fund, designed to help harden systems by reviewing code for potential vulnerabilities. The company also announced it would establish a taskforce with Australian expertise to develop policy recommendations on managing AI agent risks.
Jason Kwon, OpenAI’s chief strategy officer, is scheduled to appear before the Joint Select Committee on AI next Tuesday. Guardian Australia reported that Anthropic will also attend the hearing. Meanwhile, the federal government flagged it could introduce mandatory reporting rules for AI-related data breaches after revelations that OpenAI used a public-facing email address three months after the hack to report the incident to Services Australia. OpenAI stated it has significant work ahead to rebuild trust but is making meaningful changes.
This report was produced in the Europe Brief News newsroom.
