Executive Summary
In late 2026, researchers identified a massive surge in automated cyberattacks orchestrated through exposed dashboards of the AI-driven penetration testing tool, Cairn. Multiple activity clusters, most notably the Chinese-speaking AIA0001, have leveraged AI agents to coordinate complex attack chains, moving from reconnaissance to web shell deployment with minimal human intervention. High-profile victims include John Deere, SharkNinja, and Cummins, alongside hundreds of global e-commerce sites.
The attack campaign is characterized by the use of "agentic" workflows where AI models like DeepSeek-v4.1-Flash autonomously prioritize targets based on web traffic and verify order capabilities before attempting exploitation. Common vectors include SQL injection, insecure deserialization, and file upload vulnerabilities. The relatively low cost of these operations—estimated as low as $5 per task—signals a significant shift in attack economics, allowing adversaries to target organizations previously deemed low-value.
This trend represents a critical evolution in threat tradecraft, as AI agents can sustain operations for days, generating hundreds of unique "facts" and intents per project. The exposure of these control panels has also leaked sensitive victim data, including credentials and API keys, creating secondary risks even where exploitation failed. Organizations must prioritize the automation of their own defensive scanning and patching to match the machine-speed execution of these adversary agents.