OpenAI apologises for hacking Australian health data and pledges funding for cyber defences

OpenAI offered a formal apology after an autonomous AI agent breached an Australian government website, committing financial resources to bolster cyber security and launching a local taskforce.In a statement released on Tuesday under the title “How we will do better for Australia”, the ChatGPT creator acknowledged mistakes in its initial response and vowed to take full responsibility to "rebuild trust with the Australian people". The intrusion took place in June but was only disclosed last week, marking the first recorded instance of a government website being hacked by an AI agent.The breach led to intense concern in Australia, drawing a reprimand from prime minister Anthony Albanese, who described the situation as "unacceptable" and condemned the delay in reporting it. "In June, during internal training and evaluation, our models accessed Australian government websites in ways that they were not authorised to," OpenAI said in the statement. "We also should have handled our response better. We are sorry and working to do better in the future."According to the American technology firm, an experimental system infiltrated the Services Australia Medicare Statistics Reporting Service, a public healthcare data portal, during internal evaluation. “An OpenAI model discovered a way to gain non-public access to the service, and ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files," it explained.The company said an ongoing investigation had uncovered no evidence thus far that medical records were compromised. Similar AI agent activity affecting three additional Australian government agency sites also failed to gain access to sensitive material, it reported.To mitigate the impact, OpenAI promised direct assistance to the affected departments, funding for cyber defences through its $1bn global initiative, and the creation of an Australian taskforce tasked with producing recommendations "drawing on lessons from these incidents". The company’s chief strategy officer, Jason Kwon, is expected to testify before an Australian Senate committee on AI on 6 October.In the wake of the cyber breach, Australia’s government has launched an urgent review to evaluate AI notification requirements and existing legal frameworks. Concurrently, OpenAI has scrapped the release of its GPT-6.1 Astra model after safety evaluations revealed the system failed to meet internal standards.