A Chinese hacker used AI agents to breach over 100 corporate targets

A report by cybersecurity firm NorForge reveals that a Chinese-speaking threat actor breached the websites of over 100 corporate targets by delegating the operations almost entirely to autonomous artificial intelligence (AI) agents.

Snorre Fagerland, principal threat researcher at NorForge, uncovered the campaign while monitoring exposed instances of Cairn, an open-source AI platform. The hacker tasked AI agents using minimal instructions, comprising fewer than 100 Chinese characters, to target e-commerce and corporate platforms across the US, Canada, Australia, and Japan. Reuters independently verified that compromised targets included Deere, SharkNinja, and Cummins, though the extent of any unauthorised data extraction remains unknown.

The campaign attempted intrusions against more than 1,200 websites, deploying a mixture of DeepSeek and Anthropic's Claude Code models. Successful breaches took roughly 23 hours per site and consumed around $5,000 in computing tokens, averaging $50 per successful intrusion.

The findings provide concrete evidence that commercial autonomous AI tools can execute sophisticated cyberattacks from start to finish without direct human intervention. Fagerland also noted instances in which the agents independently mapped secondary infrastructure, identified data-processing flaws, and executed malicious server commands through sustained trial and error, marking a significant escalation in individual hacking capabilities.

Training Announcement: The BCS Foundation Certificate in AI examines the challenges and risks associated with AI projects, such as those related to privacy, transparency and potential biases in algorithms that could lead to unintended consequences. Explore the role of data, effective risk management strategies, compliance requirements, and ongoing governance of the AI lifecycle and become a certified AI Governance professional. Find out more.