Penetration-testing platforms running coordinated AI agents aren’t inherently more dangerous than those run by humans, they’re just faster. A Trend Micro analysis of a 44-agent swarm attacking Japanese organizations found that despite the sophisticated automation across reconnaissance, exploitation, and persistence, the attackers still relied on the same entry points they always have: unpatched legacy systems and default credentials.
The scale is the story. One attacker, in October 2026, probed 19 organizations with specialized agents distributing workload across the system. No breakthrough in offensive technique. No novel vulnerability class. Just the same opportunism of the 1990s now running 24/7 without human fatigue, adapted to local language and context, against thousands of targets that could be blocked by simple hygiene: patching, credential rotation, inventory of what’s actually exposed.
What made this work wasn’t AI innovation. It was the continued absence of basic security practices across targets. An automated swarm simply makes scale cost-free.
Sources: Not Only ARTEX: The Different Levels AI is Used in Attacks Targeting Japan
Coverage: CyberStrikeAI Uses 44 AI Agents to Automate Attacks Against Japan Targets • CyberStrikeAI and the Industrialized Hunt for Unlocked Doors
Related on this blog: [AUTO] 116 tech firms warn of imminent AI-powered cyberattacks • [AUTO] ClosedQuorum Outsources Command-and-Control to LLMs • [AUTO] OpenAI pauses Astra over autonomous cyber capabilities