Chinese Hacker Uses AI Agents to Breach Over 100 Corporate Websites
A threat actor leveraged exposed Carin AI agents to compromise sites for major brands including Deere and SharkNinja across multiple countries.
Key points
- A Chinese-speaking hacker used AI agents to breach over 100 websites.
- The attack targeted exposed instances of the open-source Carin platform.
- Victims include Deere, SharkNinja, and e-commerce sites in four countries.
A Chinese-speaking hacker gained unauthorized access to more than 100 corporate websites by leveraging a fleet of artificial intelligence agents, according to a report by Reuters. The campaign targeted high-profile organizations including agricultural equipment manufacturer Deere and home appliance maker SharkNinja, demonstrating the operational scale of automated threat actors.
Snorre Fagerland, principal threat researcher at NorForge, identified the activity by monitoring inadvertently exposed instances of Carin, an open-source AI problem-solving platform. Fagerland observed multiple attackers, but noted this specific Chinese-speaking actor was the most prolific in utilizing the exposed infrastructure to launch unauthorized access attempts against unsuspecting companies.
Tactics and Technical Context
The attacker commanded the AI agents using fewer than 100 Chinese characters, yet the automated systems attempted to breach more than 1,200 distinct websites by the week ending October 18, 2026. This efficiency highlights how minimal human input can drive large-scale automated campaigns when combined with capable AI tools that operate without constant supervision.
Fagerland described the tactics as sophisticated, noting that the agents did not rely on simple brute-force methods. In one documented instance, the AI scoured digital records to map target infrastructure, located a public-facing control panel, identified a specific weakness, and executed malicious commands. Fagerland told Reuters that such solutions are complex and distinguish this activity from typical script-kiddie behavior.
Scope and Impact
The campaign targeted businesses located in the United States, Canada, Australia, and Japanese e-commerce sites. While the sheer volume of attempts was high, the success rate varied among the different actors Fagerland monitored. The Chinese-speaking hacker gained unauthorized access to more than 100 sites, marking a significant breach of perimeter defenses for these organizations.
SharkNinja told Reuters that the attack only impacted a site used for product support information, limiting the potential exposure of sensitive customer data. However, other hackers observed by Fagerland had more damaging outcomes. One actor stole credit card data, while another attempted to steal funds from a cryptocurrency exchange, though the success of that specific theft remains unclear.
What should you do?
Organizations should immediately audit their internet-facing assets for exposed AI platforms or development tools like Carin that may have been left vulnerable. Security teams should monitor for unusual automated traffic patterns that indicate AI-driven reconnaissance or exploitation attempts.
What to do and how to stay safe: Carin
- Scan external networks for exposed instances of open-source AI platforms or development tools that should not be publicly accessible.
- Monitor web server logs for automated request patterns that differ from typical human browsing behavior or standard bot traffic.
- Review access controls on public-facing control panels and administrative interfaces to ensure they are not discoverable or exploitable.
- Verify that all customer-facing and internal support sites are segmented to limit lateral movement if a breach occurs.
Step-by-step guide: Social Engineering Warning Signs and How to Respond
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which companies were confirmed to be breached by this AI agent campaign?
Deere and SharkNinja were confirmed to have websites breached, with SharkNinja noting only its product support site was impacted.
What tool did the hackers use to execute these attacks?
The attackers leveraged exposed instances of Carin, an open-source AI-powered problem-solving platform, to conduct the breaches.
How did the hacker control the AI agents?
The hacker used fewer than 100 Chinese characters to command the agents, which then operated autonomously to probe and breach sites.
