• Neeve
  • Posts
  • 🧱 AI-Powered Attacks Breach Critical Infrastructure

🧱 AI-Powered Attacks Breach Critical Infrastructure

Fines, leaks, and hacks—key cyber updates

Welcome to your essential briefing on threats at the intersection of cybersecurity and critical infrastructure, brought to you by Neeve, the edge cloud security platform for smart buildings, making built spaces secure, intelligent, and sustainable.

This Week’s Cyber Insights

Jaguar Land Rover forced to shut down production at all plants after cyberattack severely disrupted retail and manufacturing operations, prompting immediate system isolation to prevent further compromise while workers were sent home from production facilities.

  • Attack discovered September 1, 2025, forcing immediate shutdown of global applications and systems to mitigate impact and prevent lateral spread across manufacturing network

  • Production facilities including Halewood Plant sent workers home, with leadership banking hours under corridor agreements while maintaining essential personnel for critical activities

  • Retail operations severely disrupted with dealers unable to register new cars, affecting customer transactions and vehicle delivery processes nationwide

  • Company working to restart global applications in controlled manner while conducting forensic investigation, currently finding no evidence of customer data theft

🤔 The Bigger Picture:

Manufacturing sector attacks demonstrate how cybersecurity incidents can immediately halt physical production and operations. Facility managers across industries must prepare incident response plans that balance operational shutdown decisions with business continuity, especially for manufacturing and industrial facilities dependent on connected systems.

A sophisticated cyberattack disrupted Nevada's state government network on August 24, forcing all state office branches to shut down operations for 48 hours through exploitation of an unpatched VPN gateway and deployment of custom malware.

  • Attack began with VPN gateway exploitation, followed by multi-stage PowerShell dropper and encrypted second-stage binary designed for privilege escalation and lateral movement

  • Custom malware established persistence through scheduled tasks masquerading as "WindowsUpdateSvc" and used WMI host process injection to evade detection

  • Threat actors deployed advanced C2 communications over HTTPS disguised as legitimate REST API calls, staging encrypted archives for data exfiltration

  • Attack disrupted email systems, public records access, and internal communications, requiring offline backup authentication services during 48-hour recovery period

🤔 The Bigger Picture:

State infrastructure attacks highlight vulnerabilities in facility management systems that rely on VPN access and legacy authentication. Building operators must assess remote access security and implement multi-layered detection to prevent similar network compromises affecting facility operations.

Cybercriminals leverage agentic AI coding assistants to conduct unprecedented automated attacks against 17 organizations across healthcare, emergency services, and government sectors, using AI for every phase from reconnaissance to ransom demand calculation.

  • Cybercriminals leverage agentic AI coding assistants to conduct unprecedented automated attacks against 17 organizations across healthcare, emergency services, and government sectors, using AI for every phase from reconnaissance to ransom demand calculation.

  • Attackers used Claude Code with structured CLAUDE.md instructions to automate network penetration, credential harvesting, and data exfiltration without traditional ransomware encryption

  • AI-powered reconnaissance scanned thousands of VPN endpoints, identified vulnerable systems, and created customized malware with anti-detection capabilities including disguised executables

  • Claude Code analyzed stolen financial data to determine ransom amounts ranging from $75,000 to $500,000, generated HTML ransom notes, and created multi-tiered extortion strategies

  • AI tools enabled low-skill actors to conduct complex operations previously requiring years of training, adapting to defensive measures in real-time

🤔 The Bigger Picture:

AI-powered attack automation fundamentally changes threat landscapes for critical infrastructure. Facility managers must prepare for AI-enhanced attacks that can rapidly adapt to security measures and conduct sophisticated operations without traditional technical expertise requirements.

Anthropic identified and disrupted sophisticated criminal operations weaponizing Claude AI for large-scale data theft, ransomware development, fraudulent employment schemes, and critical infrastructure targeting across multiple threat actor groups.

  • GTG-2002 campaign targeted 17 organizations using Claude Code for automated network penetration, data analysis, and customized extortion demands exceeding $500,000 in Bitcoin

  • UK threat actor GTG-5004 developed and marketed multiple ransomware variants with ChaCha20 encryption and anti-EDR techniques, selling on darknet forums for $400-$1,200

  • North Korean operatives used Claude for fraudulent IT worker placement schemes, creating fictitious personas and maintaining employment illusions to evade sanctions

  • Chinese actors enhanced 9-month campaign against Vietnamese critical infrastructure including telecommunications providers, government databases, and agricultural systems

🤔 The Bigger Picture:

AI tool misuse spans from individual cybercrime to nation-state operations targeting critical infrastructure. Organizations must implement AI usage monitoring and assume that threat actors have access to similar AI capabilities for attacking building automation and facility management systems.

Further Alerts & Insights

🔓 Critical Citrix 0-Day Exploited Since May in Global Attacks

CVE-2025-6543 zero-day vulnerability in Citrix NetScaler products enabled unauthenticated remote code execution against government and legal services worldwide. Attackers deployed webshells for persistent access and covered their tracks, making forensic analysis challenging despite patches released in June 2025.

🤖 Cybersecurity and AI: Managing Innovation Risks

EU AI Act requires cybersecurity leaders to catalog AI-enabled capabilities ahead of mandatory risk assessments as AI systems in enterprises increase exponentially. Organizations must balance GenAI productivity promises with content anomaly detection, data protection challenges, and AI application security risks.

🏭 CISA Releases New ICS Advisories on Critical Vulnerabilities

CISA issued three advisories for critical flaws in INVT Electric engineering tools, Schneider Electric Modicon controllers, and Danfoss refrigeration systems with CVSS scores reaching 8.7. Vulnerabilities enable remote code execution through malicious file parsing and denial-of-service via crafted FTP commands.

🌐 LLMs Transform IoT Device Communication at the Edge

New framework combines large language models with IoT networks at the edge, enabling natural language commands like "set up for movie night" to control multiple devices simultaneously. Edge-deployed LLMs reduce latency and improve privacy but introduce security concerns requiring monitoring and policy checks for physical control actions.