AI Agents Help Hacker Target 1,200 Websites and Breach More Than 100 for an Estimated $5,000
A recent cybersecurity investigation has highlighted the growing risks associated with artificial intelligence in cybercrime after a hacker reportedly used AI agents to target more than 1,200 websites and gain unauthorized access to over 100. According to a Reuters report published on October 19, 2026, the campaign involved limited human intervention, with AI agents carrying out much of the work involved in identifying potential weaknesses and attempting intrusions.
Snorre Fagerland, a principal threat researcher at cybersecurity company NorForge, observed the activity while investigating exposed installations of Cairn, an open-source AI-powered problem-solving platform. The researcher estimated that the campaign consumed approximately $5,000 worth of AI usage, making it a notable example of how automated systems may reduce the resources required to conduct large-scale cyberattacks.
The findings have raised concerns among cybersecurity professionals about the potential misuse of AI tools and the challenges businesses face in protecting websites, digital services and connected systems.
How the AI-powered campaign operated
According to the investigation, the campaign began around September 25, 2026. The operator used AI agents through Cairn to conduct attacks against websites in the United States, Canada, Australia and Japan. Many of the reported targets were e-commerce websites.
Unlike conventional attacks that may require a person to manually perform each step, AI agents can be instructed to pursue a task and continue working through multiple stages with limited supervision. In this case, the agents reportedly spent extended periods examining websites and attempting to identify exploitable weaknesses.
Fagerland said the attacker provided concise instructions outlining preferred tactics, leaving the agents to perform much of the subsequent work. By late the following week, the activity had reportedly extended to more than 1,200 targeted websites.
The investigation indicated that successful intrusions typically took around 23 hours of probing and testing. This suggests that the agents were not simply performing rapid automated scans but were also attempting more sustained investigations into potential security weaknesses.
The findings do not establish that every targeted website was successfully compromised. The reported figure of more than 1,200 refers to attempted intrusions, while the number of successful compromises was more than 100.
Estimated cost of approximately $5,000
One of the most significant aspects of the investigation was the estimated cost of the operation. Fagerland calculated that the campaign consumed approximately $5,000 worth of AI usage, equivalent to around $50 for each reported successful intrusion.
These figures are estimates based on the observed activity and the researcher's assessment of AI usage. They should not be interpreted as a verified accounting of the attacker's total expenses or as a standard price for compromising a website.
The estimate also does not measure the potential financial damage to affected companies. Costs associated with incident response, service disruption, data exposure, recovery work and legal obligations may be considerably different from the expense of running AI agents.
Nevertheless, the findings illustrate why security specialists are concerned about the economics of automated cybercrime. AI tools may enable an individual operator to attempt attacks against a much larger number of targets than would otherwise be practical with the same level of direct human effort.
John Deere, SharkNinja and Cummins among the companies identified
Reports identified websites associated with farm equipment manufacturer John Deere, consumer appliance company SharkNinja and engine manufacturer Cummins among the affected organizations.
The companies provided differing accounts of the reported incidents.
John Deere said there was no indication that sensitive information had been accessed as a result of the incident, according to published reporting. SharkNinja stated that the issue affected a standalone website containing product-support information and that the threat had been contained.
Cummins said it was investigating a report of a potential cyberattack involving a website operated by one of its third-party vendors.
These statements indicate that the reported incidents did not necessarily affect every company's systems in the same way. A compromised website does not automatically mean that a company's entire corporate network, payment infrastructure or customer database has been accessed.
Reuters reported that it could not establish the full extent of any information taken during the campaign. The amount of data accessed or stolen therefore remains uncertain based on the available reporting.
Why AI agents create new cybersecurity challenges
AI agents are software systems designed to carry out tasks using instructions, available tools and intermediate decisions. Depending on how they are configured, they can examine information, adapt their approach and continue working toward a defined objective without requiring a person to approve every step.
This capability can be valuable for legitimate activities, including software testing, vulnerability assessment, incident response and security research. However, the same capabilities may be misused when an operator directs an agent to access systems without authorization.
The NorForge findings illustrate how automated systems may continue searching for weaknesses even when an initial attempt fails. This can increase the volume and persistence of activity that defenders must monitor.
The underlying issue is not that AI tools are inherently malicious. Rather, the risk arises when powerful automation is used against systems without permission or adequate safeguards.
The investigation also highlights the importance of monitoring exposed services and third-party infrastructure. Companies frequently rely on external technology providers, support websites and connected digital platforms. Security weaknesses in these systems may create additional risks, even when the main corporate network remains protected.
Cairn platform and the role of AI tools
The campaign reportedly involved Cairn, an open-source AI platform designed for general-purpose problem-solving. The platform allows users to assign tasks to AI agents that can work through a sequence of activities.
According to the published investigation, the researcher found exposed Cairn installations associated with potentially malicious activity. More than one suspicious deployment was identified during the investigation.
The platform's creator has said that Cairn was not developed as a hacking tool and that misuse results from decisions made by individuals. The presence of a platform in an attack does not, by itself, establish that its developer supported or authorized the activity.
Reports also suggested that the operator used a combination of AI tools, including DeepSeek and Claude Code. The precise contribution of each tool and the full technical details of the campaign have not been independently established in the public reporting.
Security measures businesses should consider
The incident reinforces the importance of established cybersecurity practices as AI-powered automation becomes more widely available.
Organizations should regularly assess their internet-facing websites and applications, maintain software updates, review access permissions and monitor unusual activity. Security teams should also examine third-party services and administrative systems that may be exposed to the internet.
Companies can strengthen their defences by maintaining reliable activity logs, using multi-factor authentication where appropriate, restricting access to sensitive systems and developing clear procedures for responding to suspected intrusions.
AI can also support defensive security work by helping analysts review alerts, identify patterns and prioritize potential vulnerabilities. However, organizations should ensure that automated security tools operate within clearly defined permissions and that consequential actions remain subject to appropriate oversight.
For businesses that depend on online sales, the protection of customer information and payment-related systems is particularly important. Incident response plans should identify how affected services can be isolated, how evidence can be preserved and how customers or regulators should be notified when required.
What the findings mean for the future of cybercrime
The investigation provides an example of how AI agents may change the scale and economics of cyberattacks. A single operator reportedly used automation to attempt intrusions across a large number of websites, with a comparatively modest estimated AI usage cost.
However, the figures should be interpreted carefully. The reported number of attempted attacks does not establish the number of organizations that suffered significant damage, and the estimated cost does not represent the total cost of cybercrime or the likely cost of every future AI-assisted attack.
The consequences for each affected organization depend on which systems were accessed, whether sensitive information was exposed and how effectively the incident was contained.
The key takeaway is that AI is expanding the capabilities available to both cybersecurity professionals and malicious actors. Businesses will need to combine established security controls with stronger monitoring, regular testing and careful oversight of automated systems.
As investigations continue and more information becomes available, the full impact of this campaign may become clearer. For now, the reported incidents underline the need for organizations to prepare for cyber threats that can operate at greater speed and scale with less direct human involvement.
Source: Reuters reporting dated October 19, 2026, drawing on findings from NorForge cybersecurity researcher Snorre Fagerland.
