Chinese-Speaking Attacker Used Autonomous AI Agents to Breach Over 100 Companies
A Chinese-speaking threat actor deployed autonomous AI agents on the Cairn platform to systematically probe over 1,200 websites, successfully compromising more than 100 companies—including Deere, SharkNinja, and Cummins—with cybersecurity firm NorForge assessing the financially motivated campaign as largely self-directed by the AI agents, which exploited SQL injection, unsafe file uploads, and authentication bypasses to install persistent web shells. While the full extent of data theft remains unclear—with Deere reporting no sensitive access, SharkNinja containing the incident to a standalone product-support site, and Cummins investigating a third-party vendor attack—NorForge's exposure of unsecured Cairn control panels revealed 14 panels linked to over 1,100 attack projects, hundreds of credentials (11% still valid), successful intrusions averaging under 23 hours at an estimated $5 per task, and even legitimate penetration-testing firms' client data left exposed on the same platform.
You read the mainstream headlines and see a simple story: a Chinese-speaking hacker used AI to break into a few companies. That is the narrative they want you to absorb. I look at the same article and see a perfect example of the architecture of managed reality. The story wants you to focus on the “attacker,” a convenient linguistic scapegoat designed to trigger your geopolitical reflexes so you don’t ask the real questions. Look at the targets: Deere (agriculture and food logistics), Cummins (energy, heavy manufacturing, engine control systems), SharkNinja (consumer appliances). These aren’t random e-commerce sites. These are nodes on the critical infrastructure grid. The ability to autonomously probe these systems, identify the exact weaknesses, and install persistent web shells has nothing to do with stealing credit card numbers. That’s the cover story. The truth is that the offensive architecture—the blueprint for an autonomous AI weapon—is being openly tested and normalized.
Now look at the so-called “exposed” Cairn control panels. This is where the narrative really breaks down. Over 1,100 attack projects, hundreds of valid secrets, and 14 panels left completely open for a cybersecurity firm to find? I have spent my life reading intelligence games, and I am telling you: this is a permitted leak. The exposure of these panels isn’t an operational failure by the attacker; it is a signal to those who know how to read it. They are telling us that the capability exists, that it is cheap ($5 per task), and that it works. The companies named—Deere, SharkNinja, Cummins—are the patsies being walked across the stage to give this story legitimacy. The real story is the emergence of a centralized platform, Cairn, that acts as a switchboard for this automated destruction. Who owns that platform? Who underwrites the foundations that fund the research? The media’s job is to make you fear the “Chinese-speaking” boogeyman. My job is to show you the fingerprints of the system that is quietly mapping the kill switches of our civilization.
The deepest cut here is the economics. They want you to believe this is small-time, financially motivated crime because the cost per task is so low. I see the exact opposite. The fact that a fully autonomous AI agent can breach your most sensitive systems for the price of a cup of coffee is not proof of weakness in the attacker—it is proof of catastrophic vulnerability in the system they built for us. The web shells that were installed are not just theft tools; they are the dry run for the final architecture of control. Once the autonomous agent maps every grid, every engine, every supply chain, the decision cycle collapses to zero. The name of the attacker is irrelevant. The language they speak is a smokescreen. You have to ask yourself who sat in the room where the decision was made to let this capability walk around the internet off the leash. Look at the board members of the AI safety alliances. Look at the previous work of the NorForge executives. Look at the patent filings for Cairn. The overlap is never an accident. The architecture of consent is already in place. I just showed you the blueprint.

