A Chinese-speaking hacker used autonomous AI agents to probe more than 1,200 websites and breach over 100 commercial entities, including John Deere, SharkNinja, and Cummins, at a total operational cost of approximately $5,000, according to findings from cybersecurity firm Norforge and reporting by Reuters1,2,3.
The campaign targeted e-commerce platforms across the United States, Canada, Australia, and Japan. Norforge researcher Snorre Fagerland uncovered the operation after discovering exposed work on Cairn, an open-source AI platform that describes itself as a "general-purpose problem-solving engine". The attacker programmed agents with predefined instructions to scan network architectures and exploit uncovered weaknesses, delegating vulnerability discovery and system exploitation directly to autonomous software programs.
How the agents operated
Cairn lets users set a start and end point for AI agents, which document their progress as they move toward a task's completion. The hacker repurposed this workflow to orchestrate cyberattacks, with the AI agents carrying out breaches "with little involvement from the human hacker," according to The Verge. The agents parsed complex technical data and performed repetitive testing around the clock without manual supervision, allowing a small-scale operator to multiply reach across diverse commercial sectors.
The estimated cost per compromised company was roughly $50. Successful breaches took an average of 23 hours to compromise target architectures, and the offensive commenced on September 25, according to The420.in.
Company responses
John Deere detected no evidence indicating the extraction of sensitive corporate or customer data and stated that it instituted technical safeguards immediately. SharkNinja verified that the breach had been contained and confirmed that the targeted system held only product-support materials. Cummins disclosed that it is reviewing an intrusion directed at an online platform operated by a third-party service provider. No affected major brand reported confirmed core corporate data theft.
ANALYSIS The operation's economics are its most consequential dimension: a single actor spending $5,000 total, or roughly $50 per successful breach, achieved a scale of compromise that would have required a sizable team operating over weeks or months using manual techniques. The episode demonstrates that autonomous AI agents have materially lowered the cost and skill threshold for large-scale cyber intrusions, shifting the calculus for defenders who must now account for around-the-clock, agent-driven probing at minimal attacker expense.
The FTC recently opened formal probes into AI safety incidents at OpenAI and Anthropic[3]. ◆ The Norforge findings add a concrete, in-the-wild data point to the regulatory conversation about AI-enabled harm, illustrating offensive use of generally available agent tooling rather than frontier-model misuse.