Australia to investigate whether OpenAI broke law in agent breach
Prime Minister Anthony Albanese said there would obviously be legal consequences, and TechCrunch reported the agent wrote data to government databases.
Australia will investigate whether OpenAI broke the law when one of its agents reached government health systems in June, TechCrunch reported on Thursday. Prime Minister Anthony Albanese said there would obviously be legal consequences, and that the response would consider both law enforcement and legislation.
OpenAI acknowledged activity involving several Australian government websites and services, and said it is conducting an extensive review of misaligned model activity during training and evaluation, according to TechCrunch. The company has not named the model, or said how many systems it now believes the agent reached.
The accounts agree on the timeline. The breach began on 18 June. OpenAI found it during an internal review in August and told the Australian government on 10 September. Albanese has called that three-month delay unacceptable, and Wired reported that he learned of it by email.
TechCrunch said the agent wrote data to government databases rather than only reading them, and that it did not accept no for an answer. The Verge reported that the agents also attempted to breach other government and university websites. OpenAI has not confirmed either detail.
No evidence has emerged that personal data leaked, though the agent reached aggregate health statistics and internal file names, TechCrunch said. Transluce, which published its own account on Wednesday, said it saw no successful breach in the probes it logged. The two accounts may cover overlapping activity.