‘The threat is real’: Australia in race to stop rogue AI after model ‘escapes’ and hacks tech giant, Labor’s Andrew Charlton confirms

Australia is already testing ways to prevent artificial intelligence from going rogue, after a high-profile breach saw a model “escape” and hack a multi-billion-dollar business.  

Last week an autonomous AI model jumped the confines of a “testing environment” and hacked American tech startup Hugging Face, OpenAI confirmed. 

Experts believe the incident was a warning for more serious threats to private industry and national security to come.  

Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton told Sky News on Friday a new AI Safety Institute was racing to test rogue AI.  

“This is a global problem and models that Australians use can be created and regulated in many different countries around the world,” he said.  

“That’s why it requires a global effort.”

He said the attack, while confined to America, meant Australia and partners needed to cooperate.  

“The important thing for the Australian government is to be working with our partners in other countries to make sure that the entire global AI system is safe,” he said.  

“That’s why our AI Safety Institute … is part of a global network working with partners around the world so that in every jurisdiction that these models are subject to that same scrutiny … so that we identify these issues and stamp them out before they put Australians at risk.”  

It comes as the OpenAI security breach sparked concern from experts around the world.  

In Australia, University of New South Wales expert professor Hussein Abbass said the incident showed the threat of attacks was “real”.  

“It is no longer a question of ‘if’ AI agents go rogue and attack us by themselves,” he wrote.  

“The Hugging Face incident is an early warning that we must accelerate our preparedness. The threat is real and here. 

“More sophisticated cyber threats – the kind exemplified by the Hugging Face hack – will exploit the security layers that humans designed for human attackers, regardless of how sophisticated our designs are.” 

Australia, earlier this month, began unfurling its AI strategy as it becomes increasingly hard to ignore for governance, national security and economic prosperity.

Government and industry see an opportunity for Australia to become a global hub for the infrastructure powering artificial intelligence.

Mr Charlton told Sky News Sunday Agenda AI might be the “most important technology of our lifetimes”.

“We already have more than 200 data centres in Australia. Those data centres consume electricity equivalent to about 2 per cent of our energy grid,” he said.

“There are projections that that could grow very rapidly over the next 10 years, potentially even triple over the next 10 years.”

Mr Charlton said that it was important for Australia to become a destination for data centres in the interest of national security and sovereignty.

“This is the most important technology of our lifetimes, and we face a question as a nation about whether we want to have it in Australia,” he said.

“And if we are renting it from abroad, that means we are always paying foreigners for this technology.

“And it’s also important for security. These data centres house your photos, house your medical records. So having that data in Australia has value to us.”