AI Agents Automate Attacks on Japanese Organisations, Researchers Warn

Researchers at TrendAI, part of Trend Micro, have detailed three cyber attacks against organisations in Japan where artificial intelligence played a central role. By examining three exposed server environments used by the attackers, they found tools, scripts, target lists, logs and command histories. The setups ranged from workspaces used mostly by hand by skilled operators to AI-agent platforms that automated the whole workflow.

The first attack used CyberStrikeAI, an open-source penetration testing tool that was previously used in a campaign against Fortinet FortiGate appliances. Activity was concentrated in a short period in October 2026. The tool turned the attack process into a repeatable workflow, covering reconnaissance, vulnerability discovery and validation, exploitation, command and control, persistence, data theft and automated reporting. Work was split across 44 agents, and the platform could generate forms and fake login pages in Japanese. At least 19 organisations, including at least 16 in Japan, were subjected to validation activity, each with a report in the same format.

The second case was led by a human, with a large language model helping to speed up planning and tool development. The infrastructure suggests the campaign also aimed at AI workflow platforms and inference services, apparently to gain large amounts of inference capacity cheaply. This article covers only the opening of the research, so the third case and further detail are not included here.