Researchers have uncovered a cyber campaign targeting Japanese organizations that fully automates hacking operations using an AI framework called CyberStrikeAI. The advanced workflow leverages 44 separate AI agents to carry out reconnaissance, exploit vulnerabilities, establish persistence, exfiltrate data, and generate final reports—all with minimal human intervention. A recent investigation by TrendAI exposed this method among three distinct intrusion efforts, with this being the most fully automated. The peak activity occurred in October 2026.
Three AI-Powered Intrusion Modes
The first case illustrates the most ambitious use of AI: CyberStrikeAI was employed to run entire attack cycles end to end. It handled everything from finding targets to validating exploits and conducting command-and-control operations. Key tasks such as building fake Japanese-language login portals and formatting reports with consistent evidence indexes were all managed automatically by AI agents. At least 19 identifiable organizations were subjected to this structured workflow, 16 of them in Japan.
The second campaign blended human oversight with large language model (LLM) support. Operators used the AI to speed up both attack planning and tool creation. Victims in this category included a GIS server accessed via default credentials, from which attackers stole land registration and commercial records, email data, and Windows credentials. That operation ran for roughly 24 days between August and September 2026.
In the third pattern, AI’s role was more limited: it created scripts and tools that human operators then deployed against exposed edge devices—routers, VPNs, and servers from vendors like Cisco, Citrix, F5, and Yamaha—leveraging known vulnerabilities or credentials stored in configuration files. From January through October 2026, this actor carried out at least six campaigns, opportunistically targeting any exposed WebLogic servers or VPN endpoints it found.
Wide Reach, Language Localization, and No Confirmed Shared Operator
Though Japan was the primary target in the efforts involving CyberStrikeAI, the attacks were not entirely contained to that region. The same actor behind the first case also targeted entities in Taiwan and Hong Kong. The third pattern of attacks extended its reach to organizations in the U.K., France, Italy, Qatar, and Colombia. As for the attackers themselves, there’s no evidence that all three incidents were run by the same individual or group. The tools, operators, or the AI models used have yet to be fully identified.
Experts warn that while attack methods—like exposed services or misconfigured credentials—aren’t new, AI is dramatically changing the pace and efficiency of operations. One researcher observed that AI amplifies scale, allows tailoring to specific languages, and eliminates fatigue, enabling cyber adversaries to try thousands of attack vectors rapidly and continuously. Microsoft research shows time from vulnerability discovery to weaponization has dropped below 24 hours in many cases.
Alongside these cases, there’s been a surge in related AI-powered threats. Threat actors are increasingly integrating agentic AI systems to coordinate reconnaissance, automate social engineering, develop malware, and monetize stolen AI credentials. Frameworks like ARTEX, CARLIY0S, DeepSeek V4, and RAGE are among tools documented doing high-volume reconnaissance, credential theft, or even exploiting exposed AI endpoints.
These developments underline a stark trend: AI is no longer just an efficiency aid for threat groups—it’s becoming a core component of how cyber attacks are planned, executed, and scaled.
Why this matters: Fully autonomous or semi-autonomous AI attack tools reduce attackers’ dependence on manual effort, lowering barriers to entry and increasing speed of exploitation. Organizations must anticipate threats driven not only by new toolsets but by changing attack models—where threats are quieter, faster, and more automated. Detecting patterns—such as consistent report formats, evidence indexing, or mass targeting via AI agents—may be the key to shifting from reactive defense to proactive disruption.
