OpenAI Apologizes to Australia Over Hack of One of Its Programs Targeting Government Websites

A senior executive at OpenAI apologized before the Australian Parliament yesterday, admitting that the US technology company had mishandled the incident in which one of its independent AI programs hacked a government website last June. Jason Kwon, OpenAI’s chief strategy officer, testified at a hearing in Sydney before a parliamentary committee on artificial intelligence. “First and foremost, I would like to apologize,” Kwon said. “We are sorry. We know we have a lot of work to do to regain the trust of the Australian people, and we are committed to that.” In June, an independent OpenAI AI program breached the website of the Australian Bureau of Health Statistics and gained access to various public and private files. The model was not intended for public use. Its evaluators had removed its security protections to better assess its actual cybersecurity capabilities. OpenAI acknowledged that its program targeted four Australian government platforms, noting that it was unaware of the breach until August. The company said, “During internal training and evaluation, our models were able to access Australian government websites without being prompted. Such an incident should not have occurred. We also should have managed our response better.” Australian Prime Minister Anthony Albanese criticized the company for failing to notify the government of the incident before September.