AI Agent Swarm Breaches Over 100 Companies for $5,000; Open-Source Platform Cairn Emerges as New Attack Tool

An operator deployed an AI agent swarm through the open-source platform Cairn for about $5,000 in API costs, automatically scanning more than 1,200 targets

Starting September 25, 2026, an operator deployed an AI agent swarm through the open-source platform Cairn. Using instructions of fewer than 100 Chinese characters, the system automatically scanned more than 1,200 targets and successfully breached more than 100 enterprises, including John Deere, SharkNinja, and Cummins. The total API cost was about $5,000, averaging about $50 per successful breach, with an average time of 23 hours.

Factual Reconstruction

According to NorForge researcher Snorre Fagerland, the campaign began on September 25, 2026. The operator used agents identified as claudecode_deepseek-v4.1-flash, mainly targeting e-commerce and payment-related websites in the United States, Canada, Australia, and Japan. The agents autonomously completed the entire chain from reconnaissance and hidden administrative panel discovery to fuzzing and remote code execution, with almost no human intervention. Reuters confirmed more than 100 breached companies by reviewing an exposed panel, with a success rate of about 10%. The campaign was primarily observed because Cairn control panels were exposed.

Mechanism Breakdown

The operator provided only high-level objectives, such as implanting an uploader Web shell after code execution; the agents then used tools such as Playwright to rank targets by traffic and test them. The largest cluster, AIA0001, used Chinese-language instructions and spanned more than 1,100 projects across 14 panels. By scanning publicly exposed Cairn panels, Fagerland found about two dozen malicious instances in early October. The agents operated against a backdrop of concurrent vulnerabilities in platforms such as Warp, Railway, and LiteLLM, achieving an automated closed loop from reconnaissance to persistent access.

Industry Impact

The incident shows that open-source AI agent platforms are being used for network intrusions at scale, with the cost per successful breach falling to about $50, sharply lowering the economic barrier to traditional targeted intrusions. Among the victims, John Deere said it had taken remediation measures and found no indication that sensitive information was accessed; SharkNinja said the breach affected only a standalone product support site and that the threat was contained; Cummins said it was investigating a possible compromise of a third-party supplier website. The incident has been dubbed “Black October” in security circles, highlighting the visibility window created by exposed control planes.

Strategic Assessment

[Analysis, not fact] Open-source AI tools are empowering attackers faster than defenders can respond. If control panels for platforms such as Cairn remain exposed, similar low-cost automated attacks may spread further. Defenders need to prioritize hardening the network exposure, default credentials, and unauthenticated dashboards of agent frameworks, while strengthening Web shell detection and file integrity monitoring, to address the reality that marginal costs have fallen below the critical threshold of 1% of manual operations.