A threat actor with limited technical skills just compromised over 600 FortiGate firewalls across 55 countries. No zero-day exploit was involved. No sophisticated team behind it. They used an open-source AI platform to automate the entire kill chain — from scanning to credential theft.
If your organization has FortiGate appliances with management ports exposed to the internet, this should be at the top of your agenda today.
What happened: CyberStrikeAI and the automation of network attacks
Between January and February 2026, researchers from Team Cymru and Fortinet documented a massive campaign targeting FortiGate devices worldwide. The threat actor — identified as Russian-speaking and financially motivated — used an open-source platform called CyberStrikeAI to orchestrate attacks at scale.
CyberStrikeAI was built by a Chinese developer and published on GitHub. It integrates over 100 offensive security tools, generative AI services from multiple providers, and a full orchestration engine. In practice, it enables someone without advanced expertise to execute campaigns that previously required specialized teams and significant resources.
What makes this campaign notable is what it did not do: it did not exploit a software vulnerability in FortiGate. The attackers compromised devices through exposed management ports and weak single-factor credentials. AI did not discover a new technical flaw — it amplified basic human errors that already existed across thousands of organizations.
How AI-assisted attacks against network infrastructure work
CyberStrikeAI automates the full attack lifecycle:
Automated reconnaissance
The tool scans IP ranges looking for open FortiGate management ports (443, 8443, 10443). AI models prioritize targets based on exposure patterns and likelihood of default or weak credentials.
Intelligent credential attacks
Instead of sequential brute-force attempts, the system uses language models to generate probable credential lists based on common patterns found in enterprise network devices.
Persistence and exfiltration
Once inside, the attacker creates hidden admin accounts, modifies firewall rules, and extracts complete device configurations — including VPN credentials and security policies.
Researchers detected 21 unique IP addresses running CyberStrikeAI between January and February 2026, hosted primarily on servers in China, Singapore, and Hong Kong.
The formal assessment of the threat actor was telling: "low-to-medium baseline technical capability, significantly augmented by AI." The AI did not replace the attacker — it multiplied them.
What this changes in the threat model
Until now, large-scale attacks against network infrastructure were primarily the domain of state-sponsored APT groups or organized cybercrime networks with years of operational experience. CyberStrikeAI democratizes that capability. A single financially motivated individual with access to free AI tools can now execute campaigns that previously required six-figure budgets and teams of ten.
This is not speculation. The 21 servers detected running the tool operated for weeks before being identified. The tool's developer listed a CNNVD 2024 Vulnerability Reward Program contribution on their GitHub profile — CNNVD is operated by an entity under China's Ministry of State Security oversight.
For organizations with LATAM operations, this means the barrier to attacking your infrastructure just dropped significantly. You no longer need to be a specific target of a sophisticated group — having an exposed port in the wrong IP range is enough.
Why companies with LATAM operations are especially exposed
Latin America was among the regions directly impacted. Compromised clusters were detected across South America, the Caribbean, Central America, and other regions.
For organizations with operations in the region, three factors compound the risk:
FortiGate is the dominant firewall platform for mid-market companies in LATAM. Many organizations in Panama, Colombia, Ecuador, and Central America rely on FortiGate as their primary — sometimes only — perimeter defense.
Exposed management interfaces are more common than they should be. In our experience working with regulated companies in the region, it is not unusual to find admin interfaces reachable from the public internet — sometimes for convenience, sometimes because nobody reviewed the configuration after initial deployment.
Multi-factor authentication remains the exception, not the rule. CISA issued a specific alert regarding CVE-2026-24858, a FortiCloud SSO authentication bypass. But even without that vulnerability, devices with single-factor authentication are trivial targets for AI-automated attacks.
From a regulatory perspective, data protection laws across LATAM require organizations to implement technical safeguards proportional to risk. A firewall with default credentials or no MFA does not meet that standard, and a breach resulting from that negligence carries direct legal consequences — including fines up to $100,000 under Panama's Law 81.
What your security team should do this week
Five concrete actions your team can execute immediately:
-
Audit your FortiGate management ports. Verify that admin interfaces (HTTPS, SSH) are not accessible from the internet. If remote access is needed, restrict by source IP or require VPN.
-
Enable multi-factor authentication on all network devices. Not just FortiGate — switches, access points, and every management console. Phishing-resistant MFA (FIDO2, certificates) is ideal; TOTP is the minimum acceptable standard.
-
Review existing admin accounts. Look for accounts you don't recognize. Attackers in this campaign created hidden accounts for persistent access. Compare against your authorized inventory.
-
Update firmware to the latest stable release. Fortinet has published patches for CVE-2026-24858 and other recent vulnerabilities. A structured vulnerability management program prevents patch backlogs from becoming breach vectors.
-
Implement configuration change monitoring. If someone modifies a firewall rule or creates a new admin account, your team should know within minutes, not weeks. Early detection is the difference between a minor incident and a full breach.
Frequently asked questions
Are AI-driven attacks more dangerous than traditional ones?
Not necessarily more dangerous in technique, but far more scalable. What previously required a team of experienced attackers can now be executed by a single individual with AI tools. The volume and speed change dramatically — 600 devices across 55 countries in under 40 days.
Is my organization at risk if we use FortiGate?
FortiGate itself is not inherently insecure. The risk lies in configuration: exposed management ports, weak credentials, and lack of MFA. If your FortiGate is properly configured and updated, your exposure is significantly lower.
Do LATAM data protection laws cover this type of incident?
Yes. Laws like Panama's Law 81, Colombia's Law 1581, and Chile's Law 21.719 require proportional technical measures to protect personal data. If a breach occurs because your firewall had default credentials, regulators can determine that adequate controls were not in place — resulting in legal liability and potential fines.