Full Report
A lot of this week’s security news has the same awkward answer to one question: “Why was that allowed to work?” An extension asks for access and takes too much. A trusted service becomes part of a phishing chain. An old bug still gets results. An exposed system stays exposed. A package looks useful right up until it isn’t. Different stories, same basic problem: the path in was often already
Analysis Summary
# Morning News Roll-up September 10, 2026
## Overview
This week's threat landscape is characterized by the exploitation of inherent trust and "low-magic" entry points. Attackers are leveraging malicious browser extensions, AI-orchestrated automation for multi-stage intrusions, and the risks associated with "Shadow AI" to bypass traditional perimeters. The common theme is the exploitation of existing access paths and weak edges rather than novel zero-day vulnerabilities.
## Top Stories
### Malicious Extensions Steal Crypto Data
- Summary: A campaign involving four malicious Chrome and Firefox extensions (J7Tracker, VREO, Orbit Tracker) was identified targeting cryptocurrency traders. These extensions steal session tokens, wallet data, and Firebase access tokens, exfiltrating them to Vercel deployments.
- Source: hxxps://socket[.]dev/blog/chrome-firefox-crypto-data-theft
### AI Agents Automate Cyber Intrusions
- Summary: A Chinese-speaking threat actor is using AI frameworks (Anthropic Claude, Alibaba Qwen, DeepSeek) and the "SecFlow" orchestration tool to automate attacks against government and financial institutions across Asia and the U.S. The AI agents handle reconnaissance, exploitation, and reporting.
- Source: hxxps://hunt[.]io/blog/chinese-operator-secflow-claude-qwen-deepseek-asia
### U.K. NCSC Warns of Shadow AI Risks
- Summary: The National Cyber Security Center (NCSC) issued a warning regarding "Shadow AI"—the unauthorized use of AI tools by employees. This practice leads to intellectual property loss, data breaches, and lack of visibility into corporate information flow.
- Source: hxxps://www.ncsc[.]gov[.]uk/blogs/the-hidden-risks-of-shadow-ai
---
# Main Topic
Exploitation of Trusted Paths and AI-Driven Intrusion Automation
## Key Points
- **Weaponized Extensions:** Malicious browser extensions are being used to bypass MFA by stealing active session tokens and authenticated user info directly from the browser environment.
- **AI Orchestration:** Threat actors are moving beyond simple LLM queries to "AI orchestration frameworks" like SecFlow, which manage specialized AI agents to conduct full-spectrum cyber operations.
- **Legacy Vulnerability Re-exploitation:** AI-driven campaigns are systematically scanning for and exploiting well-known vulnerabilities (e.g., Log4Shell, Shellshock) to gain initial access.
- **Data Exfiltration via Cloud Dev Tools:** Actors are utilizing legitimate infrastructure, such as Vercel deployments, to host collection modules and receive stolen data.
## Threat Actors
- **Chinese-speaking Operator:** Attributed to recent campaigns in Asia; uses the SecFlow framework and specialized AI agents.
- **Crypto-focused Extension Publishers:** Traced back to previous malicious extensions like GhostApe and GhostApe Color.
- **Motivations:** Financial gain (crypto theft) and strategic espionage (government/financial targets in Thailand, Taiwan, and the U.S.).
## TTPs
- **AI-Driven Exploitation:** Using AI to convert campaign objectives into tasks for specialized agents.
- **Vulnerability Targeting:**
- Shellshock (CVE-2014-6271)
- Log4Shell (CVE-2021-44228)
- Spring4Shell (CVE-2022-22965)
- Shiro Deserialization and Ghostcat.
- **Post-Exploitation:** Deployment of "GLUTTON" generated web shells and "SecBox" (a Go-based remote access and network pivot framework).
- **Session Hijacking:** Stealing Firebase access tokens and application state data to bypass authentication.
## Affected Systems
- **Browsers:** Google Chrome and Mozilla Firefox (via malicious add-ons).
- **Platforms:** Axiom Trade, Padre, and users of MockApe.
- **Geographic/Sector Targets:** Government, financial, and educational systems in Afghanistan, Thailand, Taiwan, Indonesia, Vietnam, and the U.S.
- **Infrastructure:** Systems running unpatched legacy software (Nacos, Grafana, Nexus).
## Mitigations
- **Extension Controls:** Implement enterprise-wide browser management policies to restrict extension installations to approved allow-lists.
- **AI Policy:** Establish clear guidelines for "Shadow AI" usage and monitor for unauthorized AI agent activity within the network.
- **Vulnerability Management:** Prioritize patching for widely known RCE vulnerabilities like Log4Shell and Spring4Shell that AI agents are now scanning for programmatically.
- **Egress Monitoring:** Monitor for unusual outbound traffic to cloud deployment platforms like Vercel that may be used for data exfiltration.
## Conclusion
The shift toward AI-orchestrated attacks represents a significant increase in the scale and speed of intrusions. Organizations must move away from assuming that "old" bugs are no longer a threat, as automated agents can systematically probe for these weak points across vast attack surfaces. Robust identity management and strict control over browser-based access are critical to closing the "trusted paths" currently being exploited.