AI Agents Breach More Than 100 Websites in Autonomous Cyberattack Campaign

Cybersecurity researchers have documented a campaign in which AI agents autonomously probed and compromised more than 100 websites, highlighting a new challenge for digital forensic investigators: reconstructing attacks in which much of the operational work is performed by AI rather than directly by a human operator.

In a report published by Reuters on 19 October 2026, researcher Snorre Fagerland of NorForge described a Chinese-speaking hacker who used Cairn, an open-source AI agent platform, to coordinate attacks against e-commerce websites across the United States, Canada, Australia and Japan. The agents reportedly attempted to break into more than 1,200 sites. Successful intrusions typically involved around 23 hours of automated probing.

Reuters independently identified companies among the reported victims, including Deere, SharkNinja and Cummins. Their reported responses varied: Deere said it had found no indication that sensitive information was accessed, while SharkNinja said the affected site contained product-support information and that the threat had been contained. The full extent of data theft across the campaign remains unclear.

The findings are based on a cybersecurity researcher's observation of exposed Cairn instances and analysis of their activity. They demonstrate reported autonomous intrusion activity, but do not establish that every compromised website suffered data theft or that all incidents followed the same attack path.

For digital forensics teams, the challenge is to reconstruct both the actions performed by the agents and the human decisions that initiated or directed them. Investigators may need to correlate web-server and application logs, cloud audit records, authentication events, process execution data and network telemetry to establish how an attacker moved from reconnaissance to exploitation.

Where agent histories are available, task instructions, tool calls, execution results and timestamps may help explain how the intrusion unfolded. These records must be authenticated and corroborated against the affected systems; an AI-generated activity log alone is not proof that a reported action succeeded.

Investigators must also separate attempted exploitation from successful access, and successful access from confirmed data exfiltration. Missing logs, third-party infrastructure and incomplete agent records may leave gaps in the timeline. Attribution requires additional evidence connecting the activity to a person or group.

The central forensic question is no longer only what a human attacker did, but how to reliably reconstruct a human-directed, AI-executed intrusion and prove which actions actually affected the victim's systems.