OpenAI apologises for Australia health site breach

OpenAI has apologised after one of its prototype AI models bypassed safeguards and gained unauthorised access to an Australian government health statistics website.

Chief strategy officer Jason Kwon told an Australian parliamentary hearing on artificial intelligence Tuesday that OpenAI's models accessed Australian government websites “in ways they were not directed to” during internal training and evaluation.

“That should not have happened. We also should have handled our response better,” Kwon said.

The prototype sought access to the health statistics portal in June and breached a section hosting private files. OpenAI did not alert the Australian government until September, when it sent a message to a generic email inbox checked once a day.

Kwon said OpenAI knew it had work to do to rebuild trust with Australians. The apology came as the company courts Australia as a potential location for large AI data centres and is involved in building a major data centre on the outskirts of Sydney.

The hearing also covered wider concerns about AI companies, including data centre energy use and whether Australia should change copyright laws for AI training. Australian Recording Industry Association chief executive Annabelle Herd told lawmakers artists would become “roadkill” if AI companies got their way, while Anthropic spokesman Jeff Bleich said Australia risked being left behind.