Australian PM Anthony Albanese calls OpenAI AI incident 'unacceptable' after restricted data access
&imwidth=600&imheight=450&format=webp&quality=medium "Australian PM Anthony Albanese calls OpenAI AI incident 'unacceptable' after restricted data access")
Australia PM Anthony Albanese Photograph: (AFP)
OpenAI’s AI agent accessed restricted Australian government health data during research, prompting an investigation and criticism over delayed notification
Australia’s Prime Minister Anthony Albanese has described as ‘unacceptable’ an incident in which an autonomous OpenAI artificial intelligence agent accessed public and non-public files on an Australian government health statistics website. Albanese said the AI agent was conducting health-related research in June when it attempted to access a restricted batch of data. After being blocked, the system reportedly tried to find ways around the restriction. “When it was blocked, it sought ways around the blockage,” the prime minister said.
“This is a research project that has got into areas that it shouldn't have.” The incident involved an Australian government health statistics service, and the Australian government was not notified by OpenAI until September, according to Albanese. “Today I spoke with the CEO of OpenAI, Sam Altman, to express Australia's concern about this incident,” Albanese said. “And I also expressed my disappointment that it took the company way too long to inform the government what had occurred,” he said. “It took until 10 September before there was any notification at all, and the notification was an email sent to just the public mailbox.”
The prime minister said there was currently no evidence that the incident had resulted in a wider compromise of Australia's government services network. “No personal information is believed to have been accessed at this stage but investigations are ongoing,” he said. “Nonetheless this situation is obviously unacceptable.”
Australia’s cyber authorities are investigating the incident. Albanese said the Australian Signals Directorate is assisting with a forensic investigation to determine what systems may have been affected and to establish the full scope of the activity. OpenAI said its own review found that several Australian government websites and services may have been involved during an internal evaluation. “During this review, we identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation. “In the course of that, our models took actions we did not intend,” the San Francisco-based company said.
Also Read: 'There will be elections': Venezuela's interim leader Delcy Rodriguez makes big announcement at UN
The incident comes amid growing international concern about the ability of advanced AI systems to interact with computer systems and potentially bypass restrictions. OpenAI and other major AI developers have faced increasing scrutiny following several incidents involving AI models and cybersecurity testing. Earlier this year, two OpenAI models reportedly escaped a controlled testing environment and accessed internal systems at AI platform Hugging Face.
Trending Stories
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
&im=FitAndFill=(700,400))
)
)
Anthropic has also reported that its models gained unauthorized access to three unnamed organizations during testing designed to prevent interaction with real-world systems. Google has separately said its consumer AI model Gemini gained access to multiple systems by guessing login credentials. The developments have intensified calls for stronger cybersecurity safeguards as increasingly capable AI agents are given greater ability to browse websites, access information and perform tasks autonomously. Australian authorities continue to investigate the government website incident, including whether any other systems were affected.
