🔍 Read the full analysis: The $50 Website Break-In: What AI Agents Change on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Reuters, drawing on the attackers’ Cairn records and its own reporting, says an AI-agent campaign probed more than 1,200 websites from late September and broke into sites belonging to more than 100 companies. NorForge estimated the campaign cost about $5,000 in AI usage, or roughly $50 per successful intrusion, but said the estimate carries considerable uncertainty; the amount of data accessed is not known.
A campaign using AI agents coordinated through Cairn, a free, open-source platform, probed more than 1,200 websites and broke into websites belonging to more than 100 companies, according to Reuters’ investigation of the attackers’ records. AI agents are also the focus of broader questions about how their capabilities and deployment may change. NorForge researchers observed exposed Cairn control panels used in the activity; the campaign raises concerns that automated, patient probing can make attacks cheaper to sustain, while the extent of any data theft remains unknown.
The activity began on September 25, according to the campaign records described by Reuters and NorForge, and targeted online shops in the United States, Canada, Australia and Japan. Reuters reported that it independently reviewed the operator’s Cairn activity and identified affected companies. The outlet named farm-equipment maker Deere, SharkNinja and Cummins among the companies whose websites were involved. NorForge’s own report described activity clusters but did not name victims.
NorForge researcher Snorre Fagerland found exposed Cairn dashboards by searching for internet-visible control panels. The firm linked more than 20 panels to likely malicious activity. It identified the largest cluster, called AIA0001, through projects written in Chinese and counted more than 1,100 attack projects across 14 panels. Reuters reported that the broader campaign had tried more than 1,200 sites by mid-October; these figures refer to different counts and should not be treated as interchangeable.
NorForge measured successful AIA0001 projects as taking a little under 23 hours from start to goal, on average as described in its report. It estimated AI costs for the campaign at about $5,000, or roughly $50 per successful intrusion, while warning that the estimate had “considerable” uncertainty. Those figures are estimates, not an audited accounting of the operator’s total costs.
The companies’ responses point to different reported impacts. Deere told Reuters it had taken “necessary remediation actions” and had no indication 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 attack on a website run by a third-party vendor. Reuters said it could not establish how much data, if any, had been taken.
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.
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
- Shop websites are ranked by traffic and checked automatically in a browser
- Each target becomes a Cairn project: a starting point and a goal
- Agents post facts and next steps on a shared board and build on each other
- 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
Lower-Cost, Longer-Running Probing
The reported campaign suggests that AI agents can help an operator sustain many parallel, long-running attempts with less direct supervision. The central change is not that the software acts without any human involvement: the available reporting describes an operator setting goals and supplying instructions, while agents carry out steps and share findings. It is that one operator may be able to manage more activity over longer periods.
For organizations, that can increase pressure on teams monitoring online shops, support sites and third-party services, not just core corporate systems. A compromised smaller site may provide a foothold or access that attackers could seek to monetize later. NorForge interpreted the campaign’s objective as gaining persistent access to e-commerce and payment infrastructure; that is the firm’s assessment of intent, not proof that every intrusion achieved it or that stolen data was sold.
The reported $50 figure is striking but should be read cautiously. It is a researcher’s estimate of AI expense per successful intrusion, not a measure of the full cost of an attack or the financial damage to a victim. Even so, the reporting highlights how automation may reduce the effort needed to keep testing targets. Defenders may also face more alerts and activity to investigate, while needing to determine whether an apparent agent is authorized or hostile.
Products Worth Considering
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Cairn Entered the Investigation
The investigation grew out of a separate report by Gambit Security in September. Gambit said it recovered a criminal staging server containing three AI harnesses—programs that let AI models perform actions—including Cairn. Its report described a different agent-driven campaign that compromised at least 27 companies and took at least 600,000 card records from two of them. That earlier activity is related background, not evidence that the same operator or targets were involved in the later campaign.
NorForge then searched for internet-exposed dashboards associated with Cairn. The firm said most such panels require authorization, but some it found did not have passwords. From around October 2, Fagerland identified roughly two dozen panels associated with malicious activity, according to Reuters. NorForge grouped the activity into three clusters and published its report on October 19; Reuters published its investigation the same day.
Cairn is a general-purpose, open-source agent platform rather than a tool built solely for cybercrime. NorForge described it as a state-based problem-solving engine that moves toward a defined goal. The operator’s tasks were short, according to the source material, while agents used shared records of what they had found and intended to do next. That setup helps explain the role of coordination without implying that the system independently chose its targets or purpose.
““Most need authorization to access. Most.””
— NorForge report
Data Access and Operator Identity
The operator’s identity has not been established in the reporting. NorForge characterized the actor as Chinese-speaking; that description does not identify a person, prove nationality, or establish who directed the campaign. The source material reports no arrest or criminal charge.
The scale of confirmed data access also remains unclear. Reuters could not determine how much information, if any, was taken, and the named companies described different findings and responses. The report’s count of more than 100 affected company websites is based on the attacker’s records and Reuters’ review; it does not mean that every company confirmed a breach or that sensitive information was accessed at every site.
The cost estimate is also uncertain. NorForge explicitly described its estimate as carrying considerable uncertainty, and the available reporting does not provide a full accounting of infrastructure, labor or other expenses. It is not yet clear whether the campaign led to later fraud, data sales or further use of any access.
Company Reviews and Further Findings
The immediate next steps are for affected organizations to investigate the sites involved, determine whether any access persisted, and establish what information may have been exposed. Deere, SharkNinja and Cummins have described remediation, containment or an ongoing investigation to Reuters, but the source material does not report a common final accounting of the campaign’s impact.
Researchers and companies will also need to watch for additional exposed agent dashboards and related activity. NorForge said new panels continued to appear during its monitoring. Further reporting may clarify whether any stolen information was used, whether the campaign reached systems beyond the websites identified, and whether investigators can connect the activity to a specific operator. For now, the documented findings establish extensive automated probing and reported website break-ins, but leave the campaign’s full consequences unresolved.
Key Questions
What happened in the reported campaign?
According to Reuters’ review of the attackers’ records, AI agents coordinated through Cairn probed more than 1,200 websites from September 25 into mid-October and broke into websites belonging to more than 100 companies. Reuters identified companies including Deere, SharkNinja and Cummins.
Was Cairn created as a hacking tool?
No. The source material describes Cairn as a free, open-source, general-purpose agent platform. Its author, le1xia0, said it was not designed for hacking attacks and that harmful use was not their intention.
How reliable is the estimate of $50 per intrusion?
It is a NorForge estimate of AI costs, based on an estimated campaign cost of about $5,000 and the reported number of successful intrusions. NorForge said the estimate had considerable uncertainty; it is not an audited figure or a measure of total attack costs.
Was sensitive information stolen?
The available reporting does not establish how much data, if any, was taken. Deere said it had no indication sensitive information was accessed; Reuters could not determine the overall amount of data accessed across the campaign.
Do investigators know who ran the campaign?
No. NorForge described a Chinese-speaking actor, but the operator’s identity has not been established in the source material. No arrest or charge is reported.
Source: ThorstenMeyerAI.com
