Hacker Uses AI Agents to Breach John Deere and SharkNinja

Security firm Norforge has uncovered a series of cyberattacks carried out by a hacker utilizing an army of autonomous AI agents. According to reports from Reuters and The Verge, the attacker successfully breached more than 100 websites across the United States, Canada, Australia, and Japan. Among the identified victims were John Deere, SharkNinja, and Cummins.

Company responses to the breaches vary. SharkNinja stated that the attack affected a standalone site used for product support information and that the threat has been contained. John Deere reported that there is no indication that sensitive information was accessed. Cummins is currently investigating a potential attack on a website operated by one of its third-party vendors.

The attacks were orchestrated using Cairn, an open-source AI platform. Cairn describes itself as a "general-purpose problem-solving engine." The platform allows users to define a starting point and an end goal for AI agents. As these agents work toward the objective, they document their progress.

Norforge discovered the attacks by finding exposed Cairn instances. The most active attacker used the platform to automate the intrusion process with minimal human intervention. To guide the AI agents, the hacker provided "preferred intrusion tactics" written in fewer than 100 Chinese characters. Once the parameters were set, the AI agents operated largely on their own. In these cases, successful breaches were achieved in approximately 23 hours.

While the creator of Cairn, known as @le1xia0, asserts that the platform was never intended to be a hacking tool and that misuse is a human decision, the architecture of the engine makes it susceptible to weaponization. The ability for an agent to autonomously navigate toward a goal and document its path allows a user to scale attacks far beyond manual capabilities.

Norforge identified more than 20 exposed Cairn instances that they believe are malicious. This suggests that the breach of John Deere and SharkNinja was not an isolated incident but part of a broader trend of hackers leveraging open-source AI problem-solving tools to automate the discovery and exploitation of vulnerabilities.

Furthermore, the incident highlights a growing security gap. While AI tools are being developed for general productivity and problem solving, they can be repurposed to execute large-scale attacks with very little manual effort from the operator. The speed of these attacks—taking less than a day to achieve a breach—demonstrates the efficiency of autonomous agents over traditional manual hacking methods.

The use of Cairn to breach over 100 sites proves that general-purpose AI agents can be effectively weaponized to automate cyberattacks at scale.