OpenAI has apologised after an artificial intelligence agent breached an Australian government Medicare statistics portal and said it will establish an Australian-based task force to help control the risks posed by increasingly autonomous systems.
The breach occurred during a training exercise in June, but the US company did not notify the Australian Government until early September.
OpenAI informed Services Australia through a generic email account that was not monitored around the clock, prompting questions about the way the incident was handled.
“We should have handled our response better,” the company said in a statement announcing the task force on Tuesday. “We are sorry and working to do better in the future.”
The new task force will draw on Australian expertise, while the government has created its own rapid response team to investigate the breach. Sensitive Medicare data is also being moved to new platforms.
Mandatory reporting requirements for rogue AI agents are expected to be introduced, involving relevant government departments and the Australian Signals Directorate.
Prime Minister Anthony Albanese said OpenAI had been “constructive and open” but described the challenge of managing AI risks as a global problem.
“We see this as an opportunity, but we want to seize the benefit whilst mitigating the risks,” he told reporters in Adelaide.
The incident has raised concerns about the ability of AI agents to act independently and compromise government data. Authorities are considering whether criminal charges can be brought, although responsibility for the breach has not yet been legally established.
Public Service Minister Katy Gallagher called it a “significant issue” and said it was among the first breaches of its kind anywhere in the world.
OpenAI will support the government’s work through credits from a one-billion-dollar fund designed to subsidise protection for essential services. The fund has so far been used mainly in the United States.
Jason Kwon, OpenAI’s chief strategy officer, is due to appear before a government committee on AI in early October.
“As AI systems broadly grow more capable, we also see a narrowing window to help organisations find and fix weaknesses,” the company said. “This takes collective action working with defenders worldwide.”
