Censys Finds 800 Exposed Artex AI Pentest Servers as Chainalysis Hit

From unauthenticated artificial intelligence offensive frameworks left open on public cloud hosts to on-path session interception against cryptocurrency investigators, five threat disclosures this week illustrate how rapidly attackers are weaponizing developer infrastructure and edge appliances.

Censys Uncovers Over 800 Exposed Deployments of Artex AI Penetration Testing Tool

Internet intelligence firm Censys reported that its researchers identified more than 800 publicly accessible instances of Artex, an open-source AI-powered penetration testing platform that uses the Model Context Protocol (MCP) to connect Large Language Models directly to offensive utilities including nmap, sqlmap, and gobuster. Censys Principal Threat Researcher Silas Cutler warned that the exposed management panels—heavily concentrated in China and the United States on cloud providers such as Alibaba Cloud and Tencent Cloud—frequently lack authentication controls, allowing unauthorized third parties to view scan logs, harvest API keys, or launch automated reconnaissance.

The findings follow an October 30 threat intelligence advisory revealing that malicious actors weaponized Artex in two high-profile intrusions: a September 2026 compromise of a South Korean financial services company and an October 2026 intrusion into an unnamed cybersecurity vendor's ConnectWise ScreenConnect infrastructure, where attackers used the AI framework to automate lateral movement and privilege escalation.

QNAP Warns NAS Administrators to Patch "Plex2Shell" Flaws Ahead of Nov. 10 CISA Deadline

In network-attached storage security, Taiwanese hardware manufacturer QNAP issued an urgent security advisory directing administrators to update Plex Media Server to version 1.43.3.10861 or later and disable router port forwarding on port 32400 (along with unchecking Enable Relay in Plex Network settings) to block active exploitation of the "Plex2Shell" exploit chain. Discovered by security researcher Zach Main, Plex2Shell chains four distinct vulnerabilities to achieve unauthenticated remote code execution:

• CVE-2026-96652: A Server-Side Request Forgery (SSRF) flaw in the metadata agent parser allowing attackers to reach internal loopback interfaces.
• CVE-2026-96654: A directory path traversal bug exposing internal configuration files and authentication tokens.
• CVE-2026-96655: An arbitrary file write vulnerability within the transcode cache directory.
• CVE-2026-96656: An unauthenticated remote code execution (RCE) flaw triggered by executing uploaded payloads via scheduled maintenance tasks.

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-96652 and CVE-2026-96656 to its Known Exploited Vulnerabilities (KEV) catalog on October 20, setting a binding November 10, 2026 remediation deadline for federal civilian agencies.

Fake GitHub Security Repositories Infect Over 2,000 Windows Systems With AsyncRAT

Security practitioners themselves are also being targeted through trojanized open-source code. Researchers at Russian cybersecurity vendor Doctor Web (Dr. Web) uncovered a network of fraudulent GitHub repositories masquerading as penetration testing utilities, game cheats, and Windows optimization scripts that have compromised more than 2,000 Windows systems with the AsyncRAT remote access trojan.

When executed, the repositories' Python scripts download a password-protected 7-Zip archive containing a malicious loader named Microsoft.exe. To bypass User Account Control (UAC), the loader creates a counterfeit directory with a trailing space (C:\Windows \System32) and plants malicious DLLs alongside copies of trusted Windows binaries fodhelper.exe and ComputerDefaults.exe. It then neutralizes Microsoft Defender using two automated routines—Mode3, which executes PowerShell commands to register drive-wide exclusions, and Mode2, which deploys Windows Management Instrumentation (WMI) commands to disable real-time monitoring—before establishing persistence and fetching its command-and-control address (blog.masuffi[.]ru) from a dynamically updated Telegram channel.

Chainalysis Confirms On-Path Breach; 24ieye Expands AI Takedown Operations

Meanwhile, blockchain intelligence provider Chainalysis confirmed an on-path network interception breach detected between late October and early November 2026 that exposed usernames, password hashes, and active authentication tokens across all customer accounts, which include law enforcement agencies, central banks, financial regulators, and cryptocurrency exchanges. Chainalysis invalidated all active sessions, mandated password resets, and temporarily took its Reactor, Data Solutions, and Know Your Transaction (KYT) platforms offline to enforce strict IP allowlisting; internal forensics confirmed that no customer investigative graphs or proprietary case files were accessed.

Responding to the broader surge in external brand and credential targeting, cybersecurity provider 24ieye (led by Managing Director Mubarak Al Hadadi) expanded its AI-supported Digital Risk Takedown Services across North America, Europe, and the Middle East on November 7, 2026, citing Anti-Phishing Working Group (APWG) data showing 1,003,924 phishing attacks in Q1 2025 and the World Economic Forum's Global Cybersecurity Outlook 2025 finding that 72% of organizations reported increased cyber risks.