Threat reportThreat IntelligenceTL-2026-4632
CyberStrikeAI Agent Platform Abused to Automate Scanning, Exploitation, and Data Theft Against Japanese Organizations
CyberStrikeAI Agent Platform Abused to Automate Scanning (TL-2026-4632), also tracked as CyberStrikeAI abuse campaign, is a high-severity tracked intrusion set, first published 2026-10-28. It has no confirmed attribution, affects Various Legacy web systems with unauthenticated administration, maps to 8 MITRE ATT&CK techniques (T1059, T1078.001, T1190), and is covered by 9 detection rules and 10 indicators of compromise.
- Severity
- HIGHAssessed severity
- CVEs
- 0None referenced
- Techniques
- 8MITRE ATT&CK
- Actors
- 0Not attributed
- Detection rules
- 9SPL · KQL · Sigma
- IOCs
- 10Indicators of compromise
Key facts for TL-2026-4632
- Threat ID
- TL-2026-4632
- Also known as
- CyberStrikeAI abuse campaign
- Severity
- HIGH
- Status
- ACTIVE
- Category
- THREAT_INTEL
- First published
- Last reviewed
- Attribution confidence
- LOW
- Motivation
- UNKNOWN
- Target sectors
- unspecified-japanese-organizations
- Target regions
- japan, Asia-Pacific
- Detection rules
- 9
- Indicators of compromise
- 10
Malware and tooling in CyberStrikeAI Agent Platform Abused to Automate Scanning
Malware and tooling: CyberStrikeAI, MCP (Model Context Protocol) native tool orchestration, masscan, metasploit, nmap, sqlmap - S0225
How CyberStrikeAI Agent Platform Abused to Automate Scanning works
TrendAI researchers report an unidentified actor using the open-source CyberStrikeAI penetration-testing platform to coordinate 44 AI agents across reconnaissance, vulnerability discovery, exploitation, persistence, and exfiltration. At least 19 organizations were scanned, 16 of them in Japan; no confirmed breaches were reported.
TrendAI researchers Bakuei Matsukawa and Joseph C Chen reported on 2026-10-27 an October 2026 campaign in which an unidentified operator used CyberStrikeAI, an open-source, Go-based, AI-native security testing platform published on GitHub by the alias Ed1s0nZ, to orchestrate 44 specialized AI agents. The agents covered reconnaissance, vulnerability discovery, validation, exploitation, command and control, persistence, exfiltration, and reporting, and were managed through a dashboard that tracked execution success rates and token usage.
The researchers reached open directories on operator infrastructure that held scripts, target lists, logs, command histories, and attack results. They had no victim-side data. At least 19 organizations were scanned, 16 of them in Japan. Many of the recovered reports explicitly showed 'no session, no web shell, no exfiltration', so deep compromise affected only a fraction of the scanned organizations and no confirmed breaches were reported. Reported TTPs include targeting unauthenticated administration interfaces on legacy web systems, validating vulnerabilities, deploying web shells, establishing credentials for persistence, automated data theft, and low-frequency attempts distributed across many targets to evade detection. The campaign was also reported to use Japanese-language phishing content and fake login pages.
Attribution is unresolved. Simplified Chinese was used across the operator environments, but TrendAI notes that language alone does not establish location, and the activity is not confirmed as state-backed. No CVEs were cited, and the researchers withheld IPs, domains, and hashes pending investigations.
CyberStrikeAI itself (Apache 2.0, Go, 100+ YAML-defined tools, native MCP support, single- and multi-agent modes, built-in WebShell and C2 features, default port 8080 with a self-signed HTTPS web UI, SQLite storage) was earlier documented by Team Cymru as running on 21 unique IPs between 2026-01-20 and 2026-02-26 in connection with a FortiGate-targeting campaign, and ThreatBook reportedly acquired the project in October 2026. Japanese authorities and JPCERT/CC have reportedly confirmed signs of AI agents being used in attacks on Japanese companies during 2026. Severity HIGH is analyst-assigned (no CVSS). This record is based on secondary reporting because the primary TrendAI report was not retrievable.
MITRE ATT&CK techniques used in TL-2026-4632
Execution
T1059 Command and Scripting Interpreter
Persistence
T1078.001 Valid Accounts: Default Accounts; T1505.003 Server Software Component: Web Shell
Initial Access
T1190 Exploit Public-Facing Application; T1566.002 Phishing: Spearphishing Link
Resource Development
T1588.002 Obtain Capabilities: Tool
Reconnaissance
T1595.002 Active Scanning: Vulnerability Scanning; T1598 Phishing for Information
Affected products and versions in CyberStrikeAI Agent Platform Abused to Automate Scanning
- Various — Legacy web systems with unauthenticated administration interfaces
Vulnerable versions: Unspecified legacy deployments
Remediation for CyberStrikeAI Agent Platform Abused to Automate Scanning
Immediate actions
- Inventory all public-facing applications and remove or restrict unnecessary administration interfaces
- Replace default credentials on web applications and appliances
- Review VPN configurations and audit exposed AI services
- Hunt for web shells and unexpected sessions on legacy internet-facing web systems
Workarounds
- Place administration interfaces behind VPN or IP allow-lists
- Require authentication on every management endpoint
Longer-term hardening
- Patch or isolate legacy systems that cannot be updated
- Monitor for low-frequency, distributed scanning and exploitation attempts correlated across source IPs
- Add detections for AI-agent-driven tooling (scripted multi-tool scanning bursts followed by web shell upload)
Weaknesses (CWE) in CyberStrikeAI Agent Platform Abused to Automate Scanning
Timeline of CyberStrikeAI Agent Platform Abused to Automate Scanning
- Initial commit of the CyberStrikeAI GitHub repository by alias Ed1s0nZ (per Team Cymru)
- Team Cymru begins observing CyberStrikeAI deployments, later tied to a FortiGate-targeting campaign
- End of Team Cymru observation window: 21 unique IPs ran CyberStrikeAI across China, Singapore, Hong Kong, the US, Japan, and Europe
- ThreatBook reported to have acquired CyberStrikeAI and to keep an open-source edition with misuse safeguards
- Media report that a Chinese-language AI tool was likely used in cyberattacks on Japanese firms
- TrendAI (Bakuei Matsukawa, Joseph C Chen) reports an operator using CyberStrikeAI with 44 AI agents; at least 19 organizations scanned, 16 in Japan; no confirmed breaches
- Cyber Security News publishes coverage; TrendAI withholds IPs, domains, and hashes pending investigations
Sources cited for CyberStrikeAI Agent Platform Abused to Automate Scanning
- Hackers Use CyberStrikeAI Agents to Automate Network Scanning, Exploitation, and Data Theft
- Chinese AI Tool Likely Used in Cyberattacks on Japan Firms
- Open-Source CyberStrikeAI Deployed in AI-Driven FortiGate Attacks
- Tracking CyberStrikeAI Usage (Team Cymru)
- CyberStrikeAI tool adopted by hackers for AI-powered attacks
- CyberStrikeAI (Ed1s0nZ) GitHub repository
- ThreatBook acquires CyberStrikeAI security testing platform
- TrendAI Research Reinforces Japan's Urgent Cybersecurity Warning as Incidents Surge in 2026
Detection coverage for TL-2026-4632
As of 2026-10-28, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-4632 across Splunk SPL, Microsoft KQL and Sigma, covering 10 indicator(s) of compromise. The whole corpus is readable without an account; a free account unlocks full detection query text in Splunk SPL, Microsoft KQL and Sigma; paid tiers add raw indicator values, correlation and the MCP server. Threadlinqs MCP server · View plans.