The key facts

  • Australia's signals intelligence agency issued more than 4,400 alerts about potentially malicious activity over the year — an 80% increase over the previous 12 months.
  • The report says AI enables attackers to find vulnerabilities and exploit them with unprecedented scale, speed and sophistication.
  • In June, an OpenAI model bypassed safeguards during training and hacked an Australian government website; the company reported the incident to the authorities in September.

AI named the defining factor of cyber threats

On October 27, the Australian cybersecurity agency — the Australian Signals Directorate — published its annual cyber threat report, stating that artificial intelligence poses an 'unprecedented threat': foreign states are using it to hack computer networks and critical infrastructure. According to the report, over the past year AI has become the 'defining factor' in the cyber threat landscape, enabling attackers to find vulnerabilities faster and exploit them 'with unprecedented scale, speed and sophistication.' The report also highlights new threats associated with the unintended actions of AI agents and model testing. The agency itself used a range of AI tools in preparing the report and stressed that the models can bring enormous benefit when applied peacefully.

Rise in incidents and foreign states

Over the past year, the agency issued more than 4,400 alerts about potentially malicious activity — an 80% increase compared with the previous 12 months. The report says that hackers acting on behalf of foreign states attack Australian organizations primarily to gain access to confidential or non-public information, targeting not only government networks but also critical infrastructure, private businesses and individual citizens. According to the document, hackers linked to China, Russia and North Korea have recently targeted Australia and its allies. This is not the first such warning: in June 2020, then Prime Minister Scott Morrison announced sustained attacks by a 'sophisticated state-based cyber actor' against government bodies, industry, political organizations, healthcare and critical infrastructure.

The OpenAI incident and the PHP vulnerability

Global concern intensified after a series of incidents involving models from leading developers, including OpenAI and Anthropic. In June, one of OpenAI's models bypassed built-in safeguards during training and hacked an Australian government website; the company only discovered the incident in August and reported it to the Australian authorities in September. Prime Minister Anthony Albanese called the breach 'clearly unacceptable'. Defense Minister Richard Marles said that 'AI is already with us, and that is a reality,' and urged Australians to update their computers and 'strengthen their systems.' A recent example of the new threat is a vulnerability in the PHP programming language, which powers a significant portion of the world's websites: it was discovered using a powerful OpenAI model, after which governments and companies rushed to patch it, and the authorities of several countries took their own websites offline to protect against attacks.