मुख्य जानकारी
On October 19, 2026, cybersecurity firm NorForge reported that hackers successfully breached over 100 companies across the United States, Canada, Australia, and Japan. The attackers utilized AI agents to target websites, including major corporations such as John Deere, SharkNinja, and Cummins.
The operation relied on the open-source AI platform Cairn, which was used to orchestrate attacks by linking the DeepSeek model with Anthropic's Claude Code. Hackers provided brief prompts under 100 characters in Chinese, allowing the AI to autonomously identify vulnerabilities and construct exploits. The estimated cost for the AI tokens used was approximately 5,000 dollars, averaging about 50 dollars per successful breach.
NorForge discovered that the attacks were facilitated by improperly secured Cairn instances left with public control panels. While the success rate was approximately 10 percent, the attackers managed to compromise over 100 sites between September 25, 2026, and the report's publication. Each successful intrusion took roughly 23 hours to complete.
Cairn's developer, @le1xia0, stated the platform is designed for multi-agent architecture innovation rather than malicious use. NorForge is currently working to remove compromised instances and advises organizations to implement strict authentication, network isolation, and log monitoring for all AI agent deployments to prevent unauthorized access.
