Australia: “Hacking using AI is an unprecedented threat… We must prepare for disruption of core services”

(Hanoi = Yonhap News) Correspondent Park Jin-hyung = Hacking using artificial intelligence (AI) is becoming an unprecedented threat globally, and Australian authorities have warned that we must prepare for the possibility of unexpected disruption of basic core online services.

According to the AFP news agency and Australian public broadcaster ABC on the 27th (local time), Australia's cybersecurity agency, the Australian Signals Intelligence Department (ASD), announced in its annual threat report that there have been cases where AI is helping criminals and state-sponsored hackers carry out attacks on a much larger scale than before.

The report said, “Over the past year, AI has become a key element defining the cyber threat environment,” and added, “AI has enabled malicious actors to more quickly find vulnerabilities and exploit them sophisticatedly at an unprecedented scale and speed.”

According to the report, the number of cybersecurity incidents responded to by ASD in the 2025-2026 fiscal year (July 2025-June 2026) was approximately 1,500, an increase of approximately 20% from the previous year.

In particular, the number of warnings notifying the possibility of malicious cyber attacks against governments, companies, and major infrastructure surged by about 80% to about 4,400.

ASD Director Abigail Bradshaw said that the emergence of AI frontier models such as Antropic's Mythos was the most decisive event over the past year.

In this regard, the report warned that it was time to "prioritize security over system availability" and that Australians should prepare for the possibility of unexpected disruptions to basic and core online services.

“Almost every service we rely on in our daily lives today relies on digital networks in some way,” Bradshaw said. “As technological capabilities improve, it is inevitable that more service outages will occur.”

He added, “An example is a situation where websites that require patches for security must be temporarily shut down,” adding, “Chaos may appear in the form of data leaks or service interruptions.”

Previously, last June, OpenAI's AI agent caused controversy by accessing the Australian government's Medicare statistical reporting portal.

The Australian government assessed this as the first confirmed case of an AI agent gaining unauthorized access to a government system. OpenAI acknowledged that its model made unintended actions during the internal training and evaluation process and apologized to the Australian government.

Director Bradshaw said that, taking this incident as an opportunity, ASD worked with several AI companies, including OpenAI, to very quickly build an early warning protocol, and shared it with major AI research institutes and intelligence agencies belonging to the Five Eyes, a Western intelligence sharing consultative body.

The report said that foreign governments, criminals, and hackers hostile to Australia are using AI to collect information, destroy critical infrastructure, and steal financial gain, and that Australia must also respond.

Director Bradshaw emphasized the need to speed up cybersecurity modernization, saying, “To stay ahead of AI-based attackers, we must use AI tools to automate our defense systems.”