TrendAI analysed three attack infrastructures aimed at Japan and found one built on CyberStrikeAI, an open-source AI penetration testing platform. The platform split the work across 44 agents, from reconnaissance and exploitation to data theft and report writing. In October at least 19 organisations were probed, at least 16 of them in Japan, using fake login pages in Japanese. Deep compromise was limited, and the clearest intrusion came through an exposed, unprotected admin interface on a legacy system.

What it means: The research gives a realistic picture of how attackers use AI today. The big advantage is not some new, unknown technique but scale: one operator can work on dozens of targets at once, produce locally tailored fake pages and receive ready-made reports for each victim. That means smaller organisations, which used to be not worth a manual attacker's time, now end up on the list. The encouraging finding is that most attempts stopped without result, and where there was success the cause was a basic mistake: a forgotten admin interface open to the internet with no authentication. For Greek small and medium businesses, the lesson is that basic practices also protect against AI-driven attacks. It is worth taking an inventory of every service reachable from the internet, closing admin panels or putting them behind a VPN, changing default passwords and turning on multi-factor authentication. A regular scan of the company's public addresses, even with free tools, quickly shows what an attacker sees.

Read the original on TrendAI

Related service: Custom software

Related stories