Get ready for a mind-boggling revelation: a financially motivated threat actor, speaking Russian, has managed to compromise over 600 FortiGate devices across 55 countries, and here's the kicker - they did it with the help of AI!
Amazon Threat Intelligence recently uncovered this shocking activity, which took place between January 11 and February 18, 2026. CJ Moses, Chief Information Security Officer at Amazon Integrated Security, shed some light on the matter, stating that the campaign didn't exploit any FortiGate vulnerabilities. Instead, it targeted exposed management ports and weak credentials with single-factor authentication, which are fundamental security gaps that AI helped this unsophisticated actor exploit on a large scale.
The tech giant described the threat actor as having limited technical capabilities, but they overcame this constraint by relying on multiple commercial generative AI tools. These tools played a crucial role in various phases of the attack cycle, from tool development to attack planning and command generation. It's like they had their very own AI-powered toolkit!
What's even more intriguing is that the attackers used a primary AI tool as the backbone of their operation, but they also had a backup AI tool ready to assist with pivoting within compromised networks. It's like they were playing a game of chess, always having a move ready for any situation.
The threat actor is believed to be driven by financial gain and is not associated with any advanced persistent threat (APT) with state-sponsored resources. As Google recently highlighted, generative AI tools are becoming increasingly popular among threat actors, allowing them to scale and accelerate their operations, even without novel uses of the technology. In other words, AI is like a superpower that gives even novice threat actors the ability to cause serious damage.
Amazon's investigation revealed that the threat actor successfully compromised multiple organizations' Active Directory environments, extracted complete credential databases, and even targeted backup infrastructure, likely as a prelude to deploying ransomware. But here's where it gets controversial: instead of trying to persist within hardened environments or those with sophisticated security controls, the threat actor chose to move on to a softer target. It's like they used AI as a crutch to compensate for their lack of skills, making it easier to pick low-hanging fruit.
Amazon identified publicly accessible infrastructure managed by the attackers, which hosted various artifacts related to the campaign. This included AI-generated attack plans, victim configurations, and source code for custom tooling. The company described the entire operation as an "AI-powered assembly line for cybercrime." It's a chilling thought, isn't it? AI is not just a tool; it's a powerful enabler for cybercriminals.
At its core, the attacks allowed the threat actor to breach FortiGate appliances, extract full device configurations, and gain access to credentials, network topology information, and device configuration details. This involved systematic scanning of FortiGate management interfaces exposed to the internet across specific ports, followed by attempts to authenticate using commonly reused credentials. The activity was automated and sector-agnostic, indicating a mass scanning operation for vulnerable appliances. The scans originated from a specific IP address, 212.11.64[.]250.
The stolen data was then used to delve deeper into targeted networks and conduct post-exploitation activities, including vulnerability scanning using Nuclei, compromising Active Directory, harvesting credentials, and accessing backup infrastructure, which are typical steps in a ransomware operation. Amazon's data shows that the scanning activity resulted in organizational-level compromise, with multiple FortiGate devices belonging to the same entity being accessed. The compromised clusters were detected across various regions, including South Asia, Latin America, the Caribbean, West Africa, Northern Europe, and Southeast Asia.
Following VPN access to victim networks, the threat actor deployed a custom reconnaissance tool, written in both Go and Python. Analysis of the source code revealed clear indicators of AI-assisted development, such as redundant comments, simplistic architecture, naive JSON parsing, and compatibility shims for language built-ins with empty documentation stubs. It's almost as if the AI tools were doing most of the heavy lifting, leaving the threat actor with a simple, step-by-step guide to follow.
After the reconnaissance phase, the threat actor took several steps, including achieving domain compromise via DCSync attacks, moving laterally across the network, and targeting Veeam Backup & Replication servers to deploy credential harvesting tools and exploit known Veeam vulnerabilities. It's a well-coordinated attack, and AI seems to be the key enabler.
Another noteworthy finding is the threat actor's pattern of repeatedly failing when trying to exploit anything beyond the most straightforward, automated attack paths. Their own documentation recorded that the targets had either patched the services, closed the required ports, or had no vulnerable exploitation vectors. It's almost as if AI was pushing them to take the path of least resistance, and when that path was blocked, they struggled to find an alternative.
With Fortinet appliances becoming an attractive target for threat actors, organizations must ensure that management interfaces are not exposed to the internet, change default and common credentials, implement multi-factor authentication, and audit for unauthorized accounts or connections. It's also crucial to isolate backup servers, keep software up-to-date, and monitor for unintended network exposure.
As we move forward into 2026, it's essential to recognize that AI-augmented threat activity is likely to increase in volume from both skilled and unskilled adversaries. Strong defensive fundamentals, such as patch management, credential hygiene, network segmentation, and robust detection for post-exploitation indicators, remain the most effective countermeasures.
So, what do you think? Are we witnessing the beginning of a new era of AI-powered cybercrime? Will organizations be able to keep up with this evolving threat landscape? Let's discuss in the comments!
