AWS reports AI-augmented attack compromised 600+ FortiGate firewalls

Attack details from AWS incident report

AWS security teams documented a campaign from mid-January to mid-February 2026 where Russian-speaking cybercriminals compromised more than 600 FortiGate firewalls across 55 countries. The attackers used commercial generative AI tools to generate attack playbooks, scripts, and operational notes, enabling a relatively low-skilled group to run what would typically require more resources.

Attack methodology

The campaign focused on scanning for exposed FortiGate management interfaces on the public internet. Attackers then attempted commonly reused or weak credentials. Once inside, they extracted configuration files containing:

• Administrator and VPN credentials
• Network topology details
• Firewall rules

From there, they moved deeper into environments, targeting Active Directory, dumping credentials, and probing for lateral movement opportunities. Backup systems including Veeam servers were also targeted.

AI tooling characteristics

AWS observed that the AI-generated tooling was functional but rough around the edges, with simplistic parsing logic and redundant comments suggesting machine-generated code. The tools were embedded throughout the workflow rather than just used for occasional scripting. CJ Moses, CISO at Amazon, noted: "The volume and variety of custom tooling would typically indicate a well-resourced development team. Instead, a single actor or very small group generated this entire toolkit through AI-assisted development."

Attack patterns and defense

The attackers tended to abandon targets that put up resistance and move on to softer ones, emphasizing volume over finesse. Activity was geographically opportunistic rather than tightly targeted, with victims across Europe, Asia, Africa, and Latin America. Some compromises may have enabled access to managed service providers or larger shared environments, amplifying downstream risk.

AWS emphasized that basic security hygiene would have prevented most compromises:

• Keep management interfaces off the public internet
• Enforce multi-factor authentication
• Avoid password recycling

The findings follow recent warnings from Google about criminals increasingly integrating generative AI directly into operations, including using Gemini AI for reconnaissance, target profiling, phishing, and malware development.