Deere agentic AI attack — Oct 2026

What happened

According to a Reuters report published on October 19, 2026, a Chinese-speaking hacker conducted a large-scale, autonomous intrusion campaign against more than 1,200 websites across the United States, Canada, Australia, and Japan. The operation, which reportedly began on September 25, 2026, successfully breached over 100 companies, including major industrial and consumer brands such as Deere, SharkNinja, and Cummins.

Unlike traditional cyberattacks requiring significant human labor, this campaign relied on autonomous AI agents to perform reconnaissance, adapt to failed attempts, and execute exploits. The attacker utilized the open-source Cairn framework—a platform designed for general-purpose AI problem-solving—to orchestrate the agents. By chaining DeepSeek for planning and Anthropic’s Claude Code for execution, the hacker automated the entire attack loop. The operator reportedly provided fewer than 100 Chinese characters of instruction to initiate the process, which then proceeded with minimal human oversight.

Snorre Fagerland, a principal threat researcher at NorForge, identified the activity after discovering approximately two dozen exposed Cairn instances being used for malicious purposes starting around October 2. The campaign was characterized by its patience; successful intrusions typically required about 23 hours of continuous probing per target. This methodical approach allowed the attacker to operate at a remarkably low cost, with Fagerland estimating a total AI token expenditure of roughly $5,000, or approximately $50 per successful breach.

While the report confirms that the websites were breached, the full extent of data exfiltration at each company remains unclear. None of the named victims had issued public statements detailing the scope of their individual breaches as of the October 19 reporting. The incident highlights a significant shift in the threat landscape: the transition from AI-assisted hacking to autonomous, agentic operations that can scale across thousands of targets with negligible marginal cost. The creator of the Cairn platform, le1xia0, stated that the framework was not intended as a hacking tool and that its misuse was a human decision rather than a feature of the software.