How One Hacker Used AI Agents and a Few Thousand Dollars to Break Into Scores of Sites

3-line summary

• Security firm NorForge discovered malicious activity while scanning exposed instances of Cairn, a popular open-source AI problem-solving platform, used along with a mix of DeepSeek and Anthropic's Claude Code; a Chinese-speaking hacker directed attacks against more than 1,200 e-commerce sites in the US, Canada, Australia, and Japan since September 25, breaching more than 100 companies including named firms Deere, SharkNinja, and Cummins, plus two smaller firms that confirmed breaches.
• The campaign cost an estimated $5,000 in token usage, or about $50 per successful breach, with each successful intrusion typically requiring about 23 hours of AI-driven reconnaissance and vulnerability detection; other hackers used similar AI-agent tooling to steal bulk credit-card data and to attempt an attack on a cryptocurrency exchange, though the full extent of data exfiltrated was not confirmed.
• NorForge's Snorre Fagerland said this is real-world evidence that AI agents can carry out intrusions start-to-finish without human intervention, calling the techniques found 'not script kiddie stuff' and saying 'it's a new world now'; Cairn's developer said the platform was not designed for hacking and has no intent to enable abuse. Deere said it took 'necessary remediation actions' and SharkNinja said the breach was confined to an independent site containing only product-support information; similar AI-driven cybercrime has been reported in Korea and Japan, and a September campaign separately breached 25 companies and stole at least 600,000 credit-card records.

This summary is our own processing of the source article; see the source link below for full context.