OpenAI apologises for Australia hack, vows to rebuild trust

SYDNEY – OpenAI on Tuesday admitted it mishandled a hacking incident in Australia and vowed to rebuild trust in a country it is courting as a major location for data centres.

Chief strategy officer Jason Kwon appeared before Australian lawmakers to answer for an incident in which an OpenAI model bypassed safeguards and gained unauthorised access to a government health statistics website.

The incident comes amid growing concern over the ability of advanced AI systems to access computer systems without authorisation, following a series of hacking incidents involving leading AI companies, including OpenAI and rival US firm Anthropic.

“I want to begin with an apology,” Kwon told a parliamentary hearing into artificial intelligence.

“During internal training and evaluation, our models accessed Australian government websites in ways they were not directed to.

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

The OpenAI prototype sought access to a health statistics portal in June, bypassing restrictions to enter a section containing private files.

OpenAI did not alert the Australian government until September, when it sent a message to a generic email inbox that is checked only once a day.

“We are sorry. We know we have work to do to rebuild trust with the Australian people. We are committed to doing that work,” Kwon said.

With its vast renewable energy potential and stretches of uninhabited land, Australia has positioned itself as a prime location for the data centres needed to power artificial intelligence.

OpenAI is involved in the construction of a major data centre on the outskirts of Sydney.

However, AI companies also face growing opposition in Australia, with critics questioning their ethics and the environmental impact of energy-intensive data centres.

Another major point of contention is whether Australia will relax its copyright laws to allow chatbots to be trained on songs, books and other creative works.

 AI 'roadkill'

The head of the Australian Recording Industry Association told lawmakers that artists would become “roadkill” if AI companies got their way.

“There is nothing less sovereign than re-writing our laws to suit foreign commercial interests,” chief executive Annabelle Herd said.

Representatives from Anthropic, which has lobbied for copyright reform, said Australia risked falling behind.

“There is a question now for Australia – will they help set the terms on which this technology develops, or will they allow those terms to be set for them?” said company spokesman Jeff Bleich, a former US ambassador to Australia.

More than 100 organisations, including OpenAI and Anthropic, have signed an open letter calling for a global effort to strengthen defences against AI-related threats.

The letter came after two OpenAI models escaped a closed testing environment and autonomously gained access to the internal systems of Hugging Face, a platform used by AI developers to store and share code.

Anthropic recently discovered that its models had gained unauthorised access to three unidentified organisations during testing.

Google's consumer AI model Gemini also accessed multiple systems by guessing login credentials. –AFP