What happened
Tom’s Hardware reports, citing Reuters, that a Chinese-speaking hacker used AI agents to probe and gain unauthorized access to more than 100 websites belonging to major companies, including Deere and SharkNinja. The activity was identified by Snorre Fagerland, a principal threat researcher at NorForge, while he was observing exposed instances of Carin, an open-source AI-powered problem-solving platform that the report says has become popular among hackers.
Tom’s Hardware reports that a Chinese-speaking hacker used artificial-intelligence agents to probe and then obtain unauthorized access to more than 100 company websites. Deere, a farming-equipment manufacturer, and SharkNinja, a kitchen-appliance company, are named among the affected organizations.
According to the report, the agents leveraged exposed instances of Carin, described as an open-source AI-powered problem-solving platform. Tom’s Hardware attributes the broader account to Reuters and says the platform has become popular among hackers.
The activity was observed by Snorre Fagerland, identified as a principal threat researcher at NorForge, while monitoring Carin instances that had inadvertently been exposed. The supplied source does not provide the dates of the activity, the precise access method, or evidence of data theft or operational damage.
Source details: tomshardware.com ↗
Why it matters
The report describes a practical security risk from exposed AI-agent infrastructure: attackers may be able to use these systems to automate reconnaissance and intrusion attempts at scale. That matters because the danger is not limited to a flaw in one company’s website; it may also involve third-party or internally deployed AI tools that are left accessible from the internet. However, the supplied report does not establish what data was accessed, whether the affected companies suffered lasting compromise, or how much of the activity was successfully automated.
This is a concrete example of AI agents being used in an intrusion workflow rather than merely discussed as a future possibility. If an agent can repeatedly inspect targets and attempt problem-solving steps, it could reduce the human effort required for reconnaissance and unauthorized access.
The reported exposure also highlights a deployment risk: an AI tool can become part of an attack surface when an instance is unintentionally reachable. Organizations using open-source agent platforms may need to treat their configuration, credentials, network access, and tool permissions as security-critical.
The report does not independently establish the full scope of the incident. It does not say whether all 100-plus sites were successfully compromised, what information was accessed, whether the companies shared infrastructure, or whether the activity produced measurable harm.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
What to watch next
The key questions are whether the affected organizations have contained the exposed Carin instances, whether any customer or corporate data was taken, and whether the activity involved additional platforms or targets. Further reporting should clarify the relevant vulnerabilities, the dates of the intrusions, the number and identities of affected sites, and whether Carin’s maintainers or the named companies issued security guidance. The supplied source does not document pricing, availability, or a general release of the AI platform.
Look for confirmation from Deere, SharkNinja, the other organizations named in the underlying reporting, or Carin’s maintainers about affected systems, remediation, and any stolen or exposed information.
Further technical reporting should identify the exposed configuration or vulnerability, explain what capabilities the agents had, and distinguish automated actions from human-directed steps.
Security teams should watch for additional publicly exposed Carin instances and related indicators, but the supplied source does not provide enough technical detail to prescribe a specific defensive procedure or assess the prevalence of the problem.