A Hacker’s AI Agent Swarm Puts The Cost Of Break-Ins In Focus
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Hacker’s AI Agent Swarm Puts The Cost Of Break-Ins In Focus on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Reuters, drawing on records from exposed Cairn control panels, reported that an unidentified Chinese-speaking operator used AI agents to break into websites associated with more than 100 companies. NorForge estimates the campaign’s AI costs at about $5,000, but the amount of data accessed or taken has not been established.

An unidentified Chinese-speaking operator used the open-source Cairn platform to coordinate AI agents that targeted online shops, with Reuters reporting break-ins at websites associated with more than 100 companies. The campaign’s records, visible through exposed control panels, also showed attempts against more than 1,200 sites; what information, if any, was taken remains unclear.

Security firm NorForge said researcher Snorre Fagerland found Cairn control panels accessible online and observed activity across multiple systems. The firm identified more than 20 panels it linked to likely malicious use. Reuters independently reviewed activity recorded by the attacker and contacted companies it identified as victims. The operator’s identity has not been established, and the source material reports no arrest or criminal charge.

The campaign began around September 25, according to the attacker’s records described by Reuters and NorForge. By mid-October, the agents had tried more than 1,200 sites. NorForge counted more than 1,100 attack projects in the largest activity cluster, across 14 panels. Reuters named Deere, SharkNinja and Cummins among the companies whose websites were affected.

The companies described different responses. Deere told Reuters it took “necessary remediation actions” and had “no indication that sensitive information was accessed.” SharkNinja said a standalone product-support site was affected and the threat was contained. Cummins said it was investigating “a potential cyber-attack on a website operated by one of our third-party vendors.” Reuters said it could not determine how much data, if any, had been taken.

At a glance
reportWhen: Campaign activity began around Septembe…
The developmentAn investigation reported that an AI-agent campaign targeted more than 1,200 websites and recorded successful intrusions at more than 100 companies.
Thorsten Meyer AI
Cluster AIA0001 · Sep–Oct 2026

One operator, a swarm of agents

The campaign in numbers

What the researcher and Reuters report about the busiest group NorForge found. Counts come from the attackers’ own exposed panels; cost and success figures are estimates.

<100Chinese characters of instructions from the operator (Reuters)
1,200+sites the agents had tried by mid-October (Reuters)
100+companies whose websites were broken into, per Reuters' count from the attacker's records
~23haverage from start to goal for successful projects (NorForge)

The price tag estimates

NorForge / Fagerland, “uncertainty considerable”

  • ~$5 per task on average, including tasks that fail early
  • $100+ for long projects with more than 1,000 intents and facts
  • ~$5,000 in AI costs for the campaign, about $50 per successful hack
  • ~10% success rate, NorForge’s “reasonable assumption”

How the swarm runs

Simplified, no attack details

  1. Shop websites are ranked by traffic and checked automatically in a browser
  2. Each target becomes a Cairn project: a starting point and a goal
  3. Agents post facts and next steps on a shared board and build on each other
  4. Goal: a hidden way back in to shop and payment systems, for later
  • 22 SepGambit reportGambitCairn found among three AI harnesses on a criminal’s staging server (separate campaign)
  • 25 SepCampaign startsReutersA Chinese-speaking operator turns agents on shops in the US, Canada, Australia and Japan
  • ≈ 2 OctOpen panels foundReutersFagerland starts finding Cairn panels in malicious use, about two dozen
  • mid-Oct1,200+ sites triedReutersBreaks into websites of 100+ companies, incl. Deere, SharkNinja, Cummins
  • 19 OctReport and investigationNorForge · ReutersNorForge publishes “Inside Cairn”; Reuters publishes its investigation

Sources: NorForge, Inside Cairn (19 Oct 2026); Reuters (19 Oct 2026); Gambit Security (22 Sep 2026). Numbers as published, same rounding as the article.
Powered by Thorsten Meyer AI

Automation Changes the Cost Equation

The case focuses attention on how AI agents may reduce the labor needed to conduct long-running, repetitive intrusion attempts. NorForge measured successful projects in the largest cluster as taking a little under 23 hours from start to goal, with about 140 recorded facts per project. One project ran for more than five days. These figures describe the observed activity; they do not establish that every attempt succeeded or that autonomous agents can compromise any particular target.

NorForge estimated the campaign used about $5,000 in AI services, or roughly $50 per successful intrusion, while warning that the estimate carried “considerable” uncertainty. The estimate is not an independently verified accounting of the operator’s total costs. Even with that caveat, the reported scale raises a practical concern for defenders: relatively inexpensive automation can keep probing many sites, including smaller or less closely monitored services.

The reported targets included support and third-party websites, not only prominent corporate systems. Those sites can connect to a company’s broader online operations, but the available reporting does not establish that the attackers reached internal networks or obtained sensitive customer information. For companies, the incident underscores the need to account for vendor-run sites and exposed administrative tools when monitoring security.

Products Worth Considering

LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor

LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor

As an affiliate, we earn on qualifying purchases.

How Cairn Panels Exposed Activity

The investigation followed a separate September report from Gambit Security, which described another agent-driven campaign and recovered a criminal staging server containing three AI harnesses: Hermes, Strix and Cairn. Gambit reported that the separate campaign compromised at least 27 companies and took at least 600,000 card records from two of them. Those figures belong to that earlier campaign, not the activity Reuters and NorForge examined here.

NorForge then searched for publicly accessible dashboards associated with Cairn, a general-purpose, open-source system for coordinating agents. The firm said most panels required authorization, but some did not. Fagerland used internet search tools to find exposed panels and began identifying systems associated with malicious activity around October 2. NorForge grouped the activity into three clusters and published its report on October 19; Reuters published its investigation the same day.

Cairn’s design lets agents share findings and planned next steps through a common workspace, while an operator can supply instructions. NorForge describes it as a general problem-solving engine, rather than a hacking tool. Cairn’s author, who uses the name le1xia0, told Reuters the software “was not designed for hacking attacks” and said harmful use was not their intention. The platform’s general-purpose design does not establish the identity or motives of the person who used it in this campaign.

““Necessary remediation actions” were taken, and the company had “no indication that sensitive information was accessed.””

— Deere, in a statement to Reuters

Victim Impact and Costs Remain Unclear

The operator’s identity has not been confirmed. The source material describes the actor as Chinese-speaking, but that does not establish a nationality, location or identity. No arrest or charge is reported. Reuters also could not determine how much data, if any, was taken, and the named companies’ statements do not confirm a common level of access across affected sites.

The number of sites tried and the number of company websites reported as breached come from different measures: the attacker’s records showed more than 1,200 sites targeted, while Reuters counted intrusions at websites associated with more than 100 companies. NorForge’s cost estimate is approximate, and the source material does not provide enough detail to independently calculate it. The full consequences for customers, vendors and company systems remain uncertain.

Company Investigations Continue

The next developments depend on what affected companies and their vendors establish through their investigations: whether unauthorized access extended beyond the affected websites, what data may have been exposed, and whether any access remains. Cummins said it was investigating a potential attack on a third-party-operated site; Deere and SharkNinja described remediation or containment steps. The reports do not give a common timetable for those inquiries.

NorForge’s published analysis and Reuters’ reporting provide a view of activity recorded on exposed Cairn panels, but they do not answer every question about the campaign. Further confirmation of data impact, attacker attribution or law-enforcement action would change what is known. Until then, the reported counts and cost should be read with their stated limits: observations from the campaign’s records and an estimate, not a complete accounting of damage.

Key Questions

What did the AI agents reportedly do?

They were coordinated through Cairn to probe online shops and related websites. Reuters reported successful intrusions at websites associated with more than 100 companies; the operator’s records showed attempts against more than 1,200 sites.

Who operated the campaign?

The operator has not been identified. NorForge characterized the activity as coming from a Chinese-speaking actor, which does not establish the person’s identity, nationality or location.

Was customer or company data stolen?

That remains unclear. Reuters could not determine how much data, if any, was taken. Deere said it had no indication that sensitive information was accessed; the available reporting does not establish the impact across all affected sites.

How much did the campaign cost?

NorForge estimated about $5,000 in AI costs, or roughly $50 per successful intrusion, and described the uncertainty as considerable. The figure is an estimate, not a verified accounting of the operator’s total expenses.

Is this the same campaign Gambit Security reported in September?

No. Gambit Security described a separate campaign involving at least 27 companies and at least 600,000 card records taken from two of them. Those findings should not be attributed to the campaign examined by Reuters and NorForge.

Source: ThorstenMeyerAI.com

You May Also Like

Nvidia’s First GPU: GeForce 256

Nvidia released the GeForce 256 on Oct. 11, 1999, and marketed it as its first GPU, helping establish a term and product line still in use today.

Inside The Observatory’s AI Exhibit: Twelve Rooms On Future Horizons

A detailed look at The Observatory’s AI exhibit, exploring twelve key questions about AI’s impact on society, trust, jobs, and ethics, now open online.

China’s DeepSeek Debuts V4 Pro, Challenging Global AI Leaders Like Claude Fable 5

DeepSeek has reportedly launched V4 Pro, claiming performance comparable to Anthropic’s Claude Fable 5, but independent verification is not yet available.

The Role Of Cross-Domain Attacks In Compromising AI Ecosystems

Experts warn that multi-domain cyber, space, and information attacks can cascade, destabilizing AI systems and global infrastructure, with detection challenges increasing.