AI Agents Help Hacker Breach More Than 100 Organisations at an Estimated $50 per Intrusion
Autonomous AI agents helped a hacker breach websites linked to more than 100 organisations after targeting over 1,200 sites, with the campaign’s AI usage estimated at around $5,000. The activity affected companies including Deere, SharkNinja and Cummins, raising concerns about how AI tools can make large-scale cyberattacks cheaper and easier to automate.
The campaign used agents to probe websites for vulnerabilities and attempt intrusions with limited human direction. Cybersecurity researcher Snorre Fagerland of NorForge estimated that the operation consumed roughly $5,000 in AI tokens, or about $50 per successful intrusion. The estimate covers AI usage, not the full cost of carrying out an attack or the financial damage caused.
AI Agents Automated Complex Intrusion Attempts
The hacker used Cairn, an open-source AI problem-solving platform that coordinates agents to work towards a specified goal. Researchers identified exposed instances of the platform being used for malicious activity, allowing them to observe how the agents were directed and how their tasks progressed.
The campaign targeted e-commerce websites in the United States, Canada, Australia and Japan. The agents spent hours searching for weaknesses and testing possible ways to gain access. Successful intrusions typically took around 23 hours, according to Fagerland’s assessment.
The activity appeared to involve a combination of DeepSeek and Anthropic’s Claude Code. The findings do not indicate that either AI provider participated in or endorsed the attacks. Instead, they illustrate how commercially available AI tools can be repurposed by individuals to automate parts of offensive cyber operations.
Deere, SharkNinja and Cummins Respond
The reported victims experienced different levels of impact. Deere said it had taken remediation measures and had no indication that sensitive information had been accessed. SharkNinja said the incident affected a standalone website containing product-support information and that the threat had been contained.
Cummins said it was investigating a report of a potential cyberattack involving a website operated by one of its third-party vendors. These responses do not establish that the companies’ wider networks were breached or that sensitive data was stolen in every case.
The overall extent of data extraction remains unknown. A successful intrusion can provide unauthorised access without necessarily resulting in confirmed theft, and the available findings do not establish the full consequences for every affected organisation.
Low-Cost AI Automation Adds to Cybersecurity Risks
AI agents can break a complex objective into smaller tasks, test different approaches and continue working over extended periods. In cybersecurity, those capabilities can help attackers scale up vulnerability discovery and intrusion attempts without manually directing every step.
The reported campaign demonstrates how automation can reduce the amount of human effort required to pursue multiple targets. However, the estimated $50 per successful intrusion reflects the researcher’s calculation of AI usage, not a universal cost for AI-powered hacking. Actual results depend on the target’s security, the tools used and whether an attempted intrusion succeeds.
For businesses, the incident reinforces the need to monitor suspicious activity, patch vulnerable systems, secure third-party services and maintain effective incident-response procedures. Organisations should also review how exposed administrative interfaces and AI-enabled tools are secured.
As AI agents become more capable of completing multi-step tasks, the cybersecurity challenge is shifting beyond the use of AI to write code or generate phishing messages. Agents that can independently test systems and pursue access may allow individual attackers to attempt more intrusions at relatively low software cost. The extent of the risk will depend on the capabilities of the tools, the security of targeted systems and how quickly defenders detect and contain malicious activity.
