Pulse Daily — Tuesday, October 20, 2026

Edo, the headline sounds like a techno-thriller plot, but Reuters confirmed it, and Norforge documented it in detail.

A single Chinese-speaking hacker used autonomous AI agents to break into websites belonging to more than 100 companies, and the entire campaign cost an estimated $5,000 in AI token charges.

To put that in perspective, about $50 per successful intrusion.

The attacker used Cairn, an open-source AI orchestration platform, to automate the intrusion process.

Security researcher Snora Fagerland at Norforge stumbled onto the operation almost by accident while scanning the internet, He found roughly two dozen internet-exposed Cairn control panels, some with active tasks he could literally watch running.

The operator made more than 1,200 break-in attempts in total.

Successful intrusions typically took about 23 hours.

The agents just kept probing methodically.

while the human slept.

Reuters independently identified John Deere, Shark Ninja, and Cummins among the victims, spread across the U.S., Canada, Australia, and Japan.

Deere said it had remediated the issue and found no evidence sensitive data was accessed.

Shark Ninja said a standalone product support site was affected.

Cummins said it was investigating a potentially affected third-party vendor site.

And Amber, the instructions the attacker fed those agents, fewer than 100 Chinese characters, essentially a short directive, not elaborate exploit code.

That detail is the one that should sit with people.

The barrier to running a scaled cyber attack campaign used to be technical skill, time, and money.

Now it's mostly just money, and not much of it.

The economics really are as significant as the automation.

A single attacker can now attempt more than a thousand intrusions with less manual work than it would have taken to attempt ten intrusions five years ago.

The defensive implication is stark.

Security teams can no longer assume that attack volume signals a well-resourced nation state.

A motivated individual with a few thousand dollars and access to open-source orchestration tools can field what amounts to a tireless, autonomous hacking crew.

Microsoft, for its part, recognized this dynamic from the inside out.

The same day this story broke, Microsoft introduced new enterprise security controls in Microsoft 365 and Project Perception, that assess individual AI agents for risk and can restrict or shut them down before they act.

The company has explicitly warned organizations to treat AI agents as potential insider security threats, not because they are automatically malicious, but because they can be hijacked through prompt injection attacks and given legitimate system access, they then misuse.