Australia: OpenAI apologizes to government after unauthorized access to its websites

SYDNEY, Sept. 29 /ATSH-DPA/ — OpenAI has publicly apologized today after one of its artificial intelligence (AI) models gained unauthorized access to several Australian government websites.

Australian Prime Minister Anthony Albanese revealed last week that an OpenAI agent had interfered with at least one government website.

The incident had occurred in June, but the Australian authorities were notified only in September, through an email sent to a public address.

In a statement published today, OpenAI said that during internal training and evaluation processes, its models had accessed four Australian government websites in ways for which they were not authorized.

The company also admitted it had not properly managed the post-incident response. OpenAI said it wants to “rebuild the trust” of Australians and is taking steps to improve how it handles such incidents in the future.

According to the company, the case represents “a new kind of cyber incident”, pointing to an emerging global challenge as AI systems become increasingly capable of autonomously performing tasks online.

OpenAI said it would make resources and expertise available to the affected Australian agencies.

The company also plans to set up a task force in Australia that will draft recommendations on policies and ways of managing the risks associated with the use of artificial intelligence.

The company has also promised financial support for the Australian government’s cyber-defense efforts, through credits from a $1 billion fund dedicated to this purpose.

Jason Kwon, OpenAI’s chief strategy officer, is expected to appear at a hearing of an Australian government committee in early October.

The company said there is still much work ahead to restore trust and that it holds itself accountable for showing that the promised changes are actually being implemented.