Australian Prime Minister Anthony Albanese announced that an artificial intelligence (AI) agent developed by the American company OpenAI in June penetrated an Australian government health data portal, gaining unauthorized access to both public and non-public files. Reuters reported this. This is likely the first known instance of an AI agent hacking a government website.
You might also like: Lured by quick money? Pyramid schemes, also known as the ‘airplane game,’ have cost thousands their savings
Incident in Australia: OpenAI AI Agent Penetrated Medicare Database
According to Reuters, the attack represents one of the most significant incidents where AI agents have penetrated external systems outside the US, and could raise concerns about developers’ ability to keep this technology under control.
According to Albanese, the OpenAI agent gained unauthorized access to a medical statistics portal of the government agency Medicare, responsible for non-sensitive health data and statistics, including public healthcare expenditures. However, he stated that there appears to have been no unauthorized access to anyone’s personal data within the Medicare system.
“The evidence currently available suggests there has been no broader network breach. This situation is, however, obviously unacceptable, and we are expressing our extreme concern directly to OpenAI CEO Sam Altman,” the Australian Prime Minister said at a press conference in New York, where he is attending the UN General Assembly session. The incident occurred after OpenAI and Anthropic this month urged Australia to reconsider a ban that prevents them from using the country’s creative content to train their models.
Don’t miss: TradeLocker vs MetaTrader
OpenAI Admits Fault of Its AI Agents, Discloses Australian Incident With a Month’s Delay
“We observed activity involving several Australian government websites and services as our models attempted to seek answers. Our models took actions that we did not intend. Our investigation uncovered no evidence that patient records were accessed. Information accessed included aggregated health statistics and internal file names,” OpenAI stated in a press release.
Although the cyberattack occurred in June, the government only learned about it a few weeks ago. This is one of several recent incidents where OpenAI disclosed information about cyberattacks or unauthorized activity by its AI agents long after they happened. In some cases, this was because it discovered the attack late, and in others, because it initially chose not to disclose it.
Another significant incident, the intrusion into the open-source AI repository Hugging Face in mid-July, was only revealed about a week after it occurred, according to timelines published by OpenAI and independent investigators. This helped ignite a global discussion about the risks posed by increasingly powerful AI models. Competitors Anthropic, Google’s Gemini, and Meta have also reported instances where their agents accessed external systems. Some leading US AI executives, including Altman, have called for a slowdown in the development of this technology, pointing, among other things, to the threat of destructive cyberattacks from agents running out of control.
Don’t miss: Wonderinterest Trading Ltd. review
Source: ČTK
