Signs have emerged that Chinese-language hackers used AI agents to attack corporate websites in countries around the world, including the United States and Japan. As AI automatically carries out much of the process from vulnerability discovery to system intrusion, the time and cost needed for hacking are falling sharply, an analysis showed.

Japan's GigaZine reported on Oct. 20 that cybersecurity firm NorForge disclosed a case of large-scale cyberattacks that abused the open-source AI tool Cairn.

Cairn is a tool in which an AI agent explores multiple paths to solve problems when users set starting conditions and goals. It can also be used to check security vulnerabilities, but hackers abused it for actual hacking.

Snorre Fagerland (스노레 파겔란드), a senior threat researcher at NorForge, found more than 20 panels linked to malicious activity while examining Cairn management screens exposed to the internet.

The most active attack group was classified as AIA0001. The group ran more than 1,100 attack projects across 14 management screens. Its main targets were companies in the United States, Japan and Australia, and it sought in particular to secure persistent access to e-commerce and payment systems.

AIA0001 achieved its goals in 96 projects, the researchers said. Including other attack groups, there were more than 1,200 projects in total and more than 130 cases of goals achieved.

Successful projects took about 23 hours on average. AI usage costs for the overall attack activity totaled about $5,000, or about 6,623,000 won, and were estimated at about $50, or about 66,200 won, per successful case.

The affected companies cited included U.S. agricultural machinery maker Deere & Company, home appliance maker SharkNinja and engine maker Cummins.

Deere & Company said it took necessary corrective measures and found no signs of access to sensitive information. SharkNinja said it blocked threats that occurred on a separate website used for product support. Cummins is investigating the possibility of attacks on a website operated by an outside company.

Still, more than 130 cases of goals achieved do not all mean independently verified data leaks. The attackers' nationality was also not confirmed.

NorForge said the case shows hacking using AI has become automated and low-cost, and it stressed the need for companies to strengthen their security response.