Australia warns of 'unprecedented' AI threat

Artificial intelligence (AI) is creating an "unprecedented" threat, Australia's cybersecurity agency warned on Tuesday. The agency said foreign governments are using AI to attack computer networks and important infrastructure.

The Australian Signals Directorate said rapidly improving AI models could bring huge benefits if used for good. It has even used some AI tools to help prepare its own yearly report on cyber threats.

However, recent months have also shown that AI is helping criminals and hackers backed by governments carry out attacks on a bigger scale than ever before, the directorate said in the report.

"AI has been the main force shaping cyber threats over the past year. It has helped attackers find weaknesses in computer systems faster and use them on a scale, at a speed and with a level of skill never seen before," the report said.

"In recent weeks, new threats have appeared because of unexpected actions by AI agents and the testing of AI models."

Worries around the world about the power of advanced AI tools are growing after several hacking incidents involving models from leading developers, including OpenAI and Anthropic.

The report described a sharp rise in cybersecurity incidents.

Over the past year, the Australian Signals Directorate issued more than 4,400 warnings about activity that might be harmful. That was an 80% increase compared with the previous 12 months.

"Hackers backed by foreign governments continued to target Australian organizations, mainly to get access to sensitive information or information that is not public.

"These attacks are not limited to government computer networks. They also target important infrastructure, private businesses and individuals."

The report said hackers linked to China, Russia and North Korea had recently targeted Australia or its allies.

Defense Minister Richard Marles said Australia could not ignore the rapid rise of AI.

"AI is here and it's real. People need to engage with it," he told reporters.

"We cannot turn back the clock on this. Attackers are using AI to find weaknesses in computer systems."

Marles urged Australians to update their computers and make their systems harder to attack.

In recent days, governments and companies have been racing to fix serious weaknesses in a programming language used by a large share of the world's websites.

The weakness was found with the help of a powerful AI model from OpenAI.