Australian authorities have launched an in-depth investigation into a new incident involving artificial intelligence. In July, an OpenAi agent hacked an online portal of the Ministry of Health dedicated to health data, Medicare, gaining unauthorized access to the files. This is an unprecedented case in Australia, the first episode involving a government platform.
The case in Australia
A new case of violation that comes at a time when artificial intelligence is at the center of a heated debate, with several technology giants, from the CEO of Anthropic to Elon Musk, who have asked to “slow down” the development of the models. Prime Minister Anthony Albanese made the incident public: “This is an extremely worrying affair.”
The breach dates back to July 18 and affected a public portal dedicated to statistics from Medicare, the Australian national health system. According to the government, the artificial intelligence agent managed to access the site without authorization during activities that OpenAistava conducted to analyze data and information from various Australian government departments.
“The elements currently available indicate that there has not been a broader compromise of the network – explained Albanese -. I am also disappointed that the company took too long to inform the government and that the way in which the notification was made was unacceptable”.
The controversy over the times
According to the Australian government, OpenAi communicated what happened only on September 10, almost two months after the accident, through an email sent to a generic address of a government department. The US company, through a note, explained that it had discovered the episode during an internal review of the activities carried out by its models: “Our models took actions that were not our intentions”. As also confirmed by Albanese, there is no evidence of any access to personal data nor a compromise of government IT systems, but further investigations will be needed to understand how the bot managed to access without authorization.
The investigation will also look into why government systems failed to detect the breach in a timely manner. According to Deputy Prime Minister Richard Marles, this is the first case in which an artificial intelligence agent obtained unauthorized access to an Australian government system: “The impact was relatively limited, because no personal information was accessed and the system was not compromised. That said, this is a very serious incident. We are faced with an artificial intelligence agent that, inadvertently, obtained unauthorized access to a government website.”