A Hacker Breached 100+ Companies for About $5,000 Using AI Agents. That Cost Curve Just Created a New Security Audit Business.

One operator, one open-source tool, and about $5,000 was enough to breach over 100 companies. That math is the opening pitch for a new kind of security audit.

What actually happened

Between September 25 and early October 2026, a Chinese-speaking hacker ran autonomous AI agents against more than 1,200 company websites, according to Reuters reporting carried by The Next Web. The agents ran on Cairn, an open-source agent platform that NorForge principal threat researcher Snorre Fagerland found by scanning the internet for exposed deployments.

The campaign breached more than 100 companies, including farm equipment maker Deere, appliance maker SharkNinja, engine maker Cummins, and a large hardware parts vendor. Total spend across the campaign was about $5,000 in AI token costs — roughly $50 per successful breach — off a 10% average success rate. A successful break-in took under 23 hours from the first scan to final access, using web shell uploaders across at least three identified attack clusters. Hundreds of credentials were leaked in the process, with 11% verified as valid.

Fagerland's assessment: "It's not script kiddie stuff." Cairn's creator, who goes by le1xia0, responded on X: "If it is used to carry out harmful hacking attacks, that is not my intention."

Deere said it had remediated and found no indication sensitive information was accessed. SharkNinja said the incident hit an isolated product-support site and was contained. Cummins said it was investigating a potential attack involving a third-party vendor's website.

Why this creates a sellable service

The number that matters for a service business isn't 100 — it's $50. That's roughly what it now costs one operator to go from "find your site" to "breach it" using commodity, open-source agent tooling. Any company whose public-facing asset would take an attacker's agent more than a token budget's worth of effort to notice is now a target that pencils out.

That's an audit, not a penetration test: map a client's public-facing assets the way an autonomous recon agent would — exposed admin panels, stale subdomains, outdated plugins, leaked credentials sitting in old commits or paste sites — and hand over a prioritized list before an attacker's agent gets there. The pitch sells itself against the attacker's own economics: for less than it costs someone to breach one site like yours, show them which of their sites would get hit first.

The retainer layer mirrors the attack itself. This campaign ran continuously for weeks, re-scanning and retrying across clusters. A one-time audit is a snapshot; a monthly re-scan is the only way to keep pace with an attacker whose tooling never stops running.

How to apply this beyond this one campaign

The transferable fact is the cost curve, not the specific platform. Whatever agent framework hackers adopt next, the economics will look similar: cheap, parallel, continuously-running attempts against anyone with a public web presence and no one watching for it. Companies with real revenue, a public-facing e-commerce or support site, and no dedicated AppSec team — the profile hit here — describe most mid-size businesses, which makes this a wide-open addressable market for an audit service.

Sources:

https://thenextweb.com/news/ai-agents-hacker-cairn-100-sites-deere-norforge