Russian hacker uses AI tools to breach 600+ FortiGate firewalls across 55 countries in 5 weeks

Russian-Speaking Threat Actor Exploits Generative AI Tools to Scale Operations

A financially motivated Russian-speaking threat actor breached over 600 firewalls across 55 countries in just five weeks, relying heavily on commercial generative AI tools to automate and scale what would traditionally require a larger, more skilled team 1

1

2

2

. According to an incident report from AWS, the AI cyberattack campaign ran from January 11 to February 18, 2026, targeting FortiGate firewalls through exposed management interfaces and weak security credentials rather than sophisticated zero-day vulnerabilities 3

3

. CJ Moses, CISO at Amazon, described the operation as an "AI-powered assembly line for cybercrime, helping less skilled workers produce at scale" 5

5

. The volume and variety of custom tooling observed would typically indicate a well-resourced development team, yet investigators believe a single actor or very small group generated the entire toolkit through AI-assisted development 1

1

.

How the Attack Unfolded Through Automated Scanning and Brute-Force Methods

The campaign began with systematic scanning for FortiGate management interfaces exposed to the internet on ports 443, 8443, 10443, and 4443 3

3

4

4

. Rather than exploiting vulnerabilities, the attacker deployed brute-force attacks using commonly reused or weak credentials protected only by single-factor authentication 1

1

5

5

. Once inside, the attackers extracted configuration files containing SSL-VPN user credentials with recoverable passwords, administrative credentials, network topology details, and firewall rules 3

3

. These configuration files were then parsed and decrypted using AI-generated attack scripts written in Python and Go, enabling rapid lateral movement within networks 3

3

4

4

. The compromised devices were spread across South Asia, Latin America, the Caribbean, West Africa, Northern Europe, and Southeast Asia, suggesting opportunistic targeting rather than sector-specific focus 2

2

3

3

.

AI-Generated Code Reveals Lowering the Barrier for Cybercrime

Analysis of the source code revealed clear indicators of AI-assisted development, including redundant comments that merely restate function names, simplistic architecture with disproportionate investment in formatting over functionality, and naive JSON parsing via string matching rather than proper deserialization 3

3

4

4

. While functional for the threat actor's specific use case, the tooling lacked robustness and failed under edge cases—characteristics typical of AI-generated code used without significant refinement 3

3

. The attacker utilized at least two large language model providers throughout the campaign to generate reconnaissance tools, develop credential extraction utilities, and create operational documentation 3

3

. In one instance, the actor submitted a full internal victim network topology, including IP addresses, hostnames, credentials, and known services, to an AI service and requested help spreading further into the network 3

3

. This demonstrates how generative AI tools are fundamentally changing the threat landscape by enabling less sophisticated actors to conduct operations previously reserved for well-resourced teams.

Targeting Backup Infrastructure Signals Ransomware Intent

Following VPN access to victim networks, the threat actor deployed custom reconnaissance tools to analyze routing tables, classify networks by size, run port scans, identify SMB hosts and domain controllers, and search for HTTP services 3

3

. The campaign specifically targeted Active Directory environments, using tools like Meterpreter and mimikatz to conduct DCSync attacks against Windows domain controllers and extract NTLM password hashes 3

3

. Notably, the attacker showed particular interest in Veeam Backup & Replication servers, deploying custom PowerShell scripts and compiled credential-extraction tools while attempting to exploit known Veeam vulnerabilities including CVE-2023-27532 and CVE-2024-40711 3

3

. Security experts note that threat actors typically target backup infrastructure before deploying ransomware to prevent restoration of encrypted files from backups, suggesting the campaign was setting up for ransomware attacks 2

2

3

3

. When encountering more hardened security environments, the attacker simply moved on to easier targets rather than persisting, reinforcing that volume rather than technical sophistication was the winning strategy 1

1

2

2

.

Basic Cybersecurity Hygiene Could Have Prevented Most Breaches

AWS emphasizes that basic cybersecurity hygiene measures would have shut down much of the activity before it gained traction 1

1

. Recommendations include ensuring FortiGate management interfaces are not exposed to the internet, enforcing multi-factor authentication, implementing unique complex passwords for all accounts, and ensuring VPN passwords differ from Active Directory credentials 3

3

5

5

. Moses warns that organizations should anticipate AI-augmented threat activity will continue to grow in volume from both skilled and unskilled adversaries 2

2

. The findings arrive weeks after Google warned that criminals are increasingly integrating generative AI directly into their operations, including abuse of its Gemini AI chatbot for tasks ranging from reconnaissance and target profiling to phishing and malware development 1

1

3

3

. Last year, Anthropic reported that a hacker leveraged its technology as part of a vast cybercrime scheme impacting at least 17 organizations, marking what was then an unprecedented instance of attackers weaponizing commercial artificial intelligence tools on a widespread basis 2

2

.