Lone Hacker Compromises Over 100 Websites Using Autonomous AI Agents

A single hacker has successfully compromised more than 100 websites belonging to businesses across the United States, Canada, Australia, and Japan over the past few weeks. According to a recent report published Monday by cybersecurity firm NorForge, the attacker utilized a network of autonomous AI agents rather than traditional hacking collectives or large syndicates. The campaign targeted lax cybersecurity systems to steal sensitive data, credit card numbers, and cryptocurrency.

The unnamed Chinese-speaking hacker coordinated the operation using Cairn, an open-source AI problem-solving and state-management engine available on GitHub. Cairn allows users to organize AI agents and assign them tasks to locate vulnerabilities and breach systems. Security researchers discovered that many hackers left their Cairn panels exposed, enabling firms like NorForge to monitor prompt-and-response actions in real time. The underlying artificial intelligence models powering the operation consisted largely of a mix of China's DeepSeek and Anthropic's Claude Code.

Starting September 25, the automated campaign attempted to breach more than 1,200 sites, achieving a success rate of roughly 10 percent. The entire multi-site hacking effort cost approximately $5,000 in API tokens, amounting to roughly $50 per successful breach. Security researchers noted that the prompts required minimal technical skill, with some instructions consisting of fewer than 100 Chinese characters or simple commands like fetching credit cards and cryptocurrency. The automated agents averaged about 23 hours to complete a successful intrusion, which included scanning digital records, locating control panels, and crafting exploits to execute malicious commands.

The creator of Cairn, posting under the handle @le1xia0 on Twitter, addressed the misuse of the platform by stating that the software is an innovative multi-agent architecture not intended for malicious cyberattacks. Despite these disclaimers, the incident highlights a growing trend where low-cost, automated AI tools enable single operators to scale cyber attacks that previously required coordinated groups of individuals.