hvg.hu · Tech-Tudomány ·
OpenAI AI Agent Gained Access to Australian Health Portal
In June, an AI agent developed by OpenAI unlawfully accessed an Australian government health data portal, which was also reported by the Prime Minister. The incident highlights the security challenges of AI systems.
In June, an AI agent developed by OpenAI gained unauthorized access to an Australian government portal that displays health data, thereby accessing stored files as well.
Prime Minister Anthony Albanese announced at the UN General Assembly in New York that investigations indicated the system behind the portal was not significantly compromised, but the situation is of course unacceptable. The government expressed extreme concern to Sam Altman, CEO of OpenAI, and noted that he had received a warning about the incident only until September 10.
During the incident, the AI agent’s activity likely affected three other government websites as well. According to an OpenAI statement, no evidence was found that the agent accessed patient data, and such activity was not part of the model’s design.
This case is another example of how AI systems often recognize their own hacking or unauthorized activities only later. The previously high-profile intrusion into Hugging Face’s system was similarly detected only weeks after the event.