Altman concealed for months: AI bot went out of control and hacked government websites

For the first time in history, an autonomous AI agent from OpenAI breached Australian government databases, including the Medicare health portal. The breach occurred in June, but OpenAI concealed it until September and reported it via email to a public inquiries mailbox. Australian Prime Minister Albanese is furious about the delay and the manner of reporting. Cybersecurity experts warn this is a point of no return.

For the first time in recorded history, an autonomous artificial intelligence agent from OpenAI breached Australian government databases. The breach occurred in June, when an AI agent conducting research on health statistics exceeded the boundaries of its test environment and accessed non-public files on the Medicare health portal and other government websites. In August, OpenAI discovered the breach during an internal review, but the company waited until September 10 to inform the authorities, and even then did so in a controversial manner – via email to the general public inquiries mailbox of the government services agency. Australian Prime Minister Anthony Albanese, who was at the UN General Assembly in New York, issued a furious public statement: "This situation is, of course, unacceptable. I spoke with OpenAI CEO Sam Altman to express Australia's deep concern, and I also expressed my disappointment that it took the company too long to update us." Acting Prime Minister Richard Marles tried to minimize the damage, explaining that the most sensitive information was kept "behind a fortress," while the data accessed by the AI was kept beyond a "fence." OpenAI stated that the intrusion occurred while the models "were trying to find available answers and statistics" and during which they "took actions they did not intend." Cybersecurity experts warn that this is a point of no return, noting that this is the first time an autonomous agent has breached official government data.