Skip to content
Sign in
Defense & Security · Global

Reuters reports AI-agent hacker breached John Deere, SharkNinja and Cummins

A hacker used AI agents to breach more than 100 websites, including sites associated with John Deere, SharkNinja and Cummins. John Deere reported no indication that sensitive information was accessed, while SharkNinja said the affected site held product-support information and the threat had been contained.

AI-assisted synthesis · Based on 4 outlets' reporting · Article text read from 2 of 4 outletsHow it's made

View timeline · 8 entries
4 outletsLatest reporting 3 min read
Coverage: Center OnlyL 0 · C 3 · R 0 · 1 unrated

What happened

Key facts

  • John Deere and SharkNinja were among the companies whose websites were breached (per The Verge, martincid.com, qz.com).
  • Cummins was among the companies whose website was breached (per The Verge, martincid.com).
  • Norforge said the AI agents operated mostly autonomously (per The Verge).
  • The hacker found about two dozen exposed Cairn setups running malicious jobs (per martincid.com, The Verge).
  • John Deere said it had no indication that sensitive information was accessed (per The Verge, martincid.com).
  • SharkNinja said the breach affected a standalone website containing product-support information and that the threat had been contained (per martincid.com).

The brief

A hacker used AI agents to breach more than 100 websites, including sites associated with John Deere, SharkNinja and Cummins. Norforge, the AI security firm that identified the activity, said the agents largely operated on their own. The affected companies named in the reporting span farm equipment, appliances and engines.

Norforge found more than 20 exposed Cairn instances it characterized as likely malicious. Martincid.com reports that the hacker located about two dozen exposed Cairn setups running malicious jobs and could read the agents’ work logs. These accounts describe exposed systems as part of how the activity was observed; they do not establish how many companies were affected in total.

The reported activity involved companies in the United States, Canada, Australia and Japan. The Verge says the hacker was the most active among the exposed Cairn instances Norforge identified and that the campaign targeted companies in those four countries. Martincid.com describes the target set as e-commerce sites in the same countries.

The sources therefore differ in emphasis—companies broadly versus e-commerce sites—while naming the same countries. John Deere told Reuters there was no indication that sensitive information had been accessed and said it had taken remediation actions.

SharkNinja said the breach involved a standalone site containing product-support information and that it had contained the threat. Cummins said it was investigating a report of a potential cyberattack on a website operated by a third-party vendor.

Those statements describe different stages and scopes: John Deere reported no sign of sensitive-data access, SharkNinja described containment, and Cummins said its investigation was ongoing. The Verge reports that the hacker described preferred intrusion tactics in fewer than 100 Chinese characters and that successful breaches took about 23 hours.

These details suggest the hacker supplied concise instructions to the agents, while Norforge’s finding that they mostly operated autonomously indicates that much of the execution was automated. The reporting does not specify whether those timing and instruction details apply to every breach.

Cairn’s creator, who goes by @le1xia0, said the platform was not designed as a hacking tool and described misuse as a human decision rather than a feature of the system. Norforge said it was working to take the malicious Cairn setups offline.

The reported account links the incidents to misuse of exposed systems, while the platform creator places responsibility on users who deploy the software.

Why it matters

  • John Deere, SharkNinja and Cummins face direct costs from remediation, threat containment or investigation after breaches involving their websites.
  • SharkNinja said a standalone product-support site was affected, placing the availability and integrity of that support information at stake; it said the threat was contained.
  • John Deere reported no indication that sensitive information was accessed, while Cummins was still investigating a potential attack through a third-party vendor’s website.
  • The reported target geography spans the United States, Canada, Australia and Japan, and Norforge’s findings connect the activity to exposed Cairn instances running malicious jobs.

What to watch next

  • Whether Norforge succeeds in taking the exposed Cairn setups it considers malicious offline.
  • Whether Cummins’ investigation confirms a cyberattack on its third-party vendor’s website.
  • Whether John Deere identifies any evidence of sensitive information access after its remediation actions.
  • Whether SharkNinja reports further consequences after saying it contained the threat to its product-support website.

How it's being covered

Coverage analysisShow details
4 source articles · 4 unique outlets
Outlet political mix: Center Only· 3 Center · 1 unrated
Where the outlets are based: US/Canada 50% · Global 25% · Other 25%
Outlet political mix
Center Only
Left
Center
Right
0 outlets
3 outlets
0 outlets
The marker shows coverage balance: each unique outlet counts once. 1 unrated outlet is not included.
Geography
Where the outlets covering this story are based.
4 unique outletsDominant coverage: US/Canada
Coverage imbalance
  • Coverage is currently concentrated among Center-rated outlets (3 of 3 rated sources); comparatively few Left or Right-rated outlets have published on the event.
How we measure coverage

One source is from a wire service (AP, Reuters, AFP or Bloomberg), whose reporting other outlets often republish.

How outlets' accounts compare

1 framing difference · 1 uncorroborated claim · 2 information gaps
The reports repeat the same account with different emphasis; its central claims are not independently confirmed in the reporting Updat3 read.Compared across 2 reports whose article text Updat3 read; 2 other outlets were checked by headline only.
Framing / emphasis differences
Sources emphasise different angles, actors or consequences of the same reporting.
  • The Verge emphasizes largely autonomous AI agents and describes the targets as companies; Martincid.com emphasizes exposed Cairn setups and characterizes the campaign’s targets as e-commerce sites.
Updat3: These are differences of emphasis, not disputes about what is reported. They do not show that the reported claims are true.
One-sided or uncorroborated claims
Claims that come from one party or one source family and are not independently corroborated.
  • The Cairn creator, @le1xia0, said the platform was not a hacking tool and characterized misuse as a human decision.
Updat3: Not independently corroborated yet — treat as a claim, not an established fact.
Not established by the reporting
Information the available reporting does not contain or confirm. This is a gap, not a disagreement.
  • The sources do not describe direct consumer losses or provide a count of customers affected.
  • The sources do not establish how many companies were affected or provide a company-by-company breakdown of the more than 100 breached websites.
Updat3: Updat3 cannot conclude anything about this from the current sources.

Timeline

Developments from this story's sources, dated background we compiled and fact-checked, and our earlier coverage. Each row says which.
  1. The US Computer Fraud and Abuse Act, enacted in 1986, criminalized unauthorized access to protected computers and became a central federal law for prosecuting computer intrusions.
    Background · from our historical context
  2. In November 1988, the Morris worm spread across connected computers, demonstrating how automated code could exploit networked systems at scale.
    Background · from our historical context
  3. Mirai malware infected internet-connected devices and used them in distributed denial-of-service attacks.
    Background · compiled and fact-checked by Updat3
  4. In January 2023, the US National Institute of Standards and Technology released its AI Risk Management Framework, a voluntary guide for identifying and managing risks in AI systems.
    Background · from our historical context
  5. In June 2023, the Open Worldwide Application Security Project published its first Top 10 list of security risks for large-language-model applications, including prompt injection and insecure handling of model outputs.
    Background · from our historical context
  6. OpenAI had introduced function calling for GPT models in June 2023, enabling models to request that software execute specified operations.
    Background · from our historical context
  7. The US Cybersecurity and Infrastructure Security Agency and partner agencies published secure-by-design guidance for technology manufacturers.
    Background · compiled and fact-checked by Updat3
  8. On October 22, 2024, Anthropic introduced a computer-use capability for its Claude model, allowing it to interpret screen images and control a computer through simulated mouse and keyboard actions.
    Background · from our historical context

How we got here

On October 22, 2024, Anthropic introduced a computer-use capability for its Claude model, allowing it to interpret screen images and control a computer through simulated mouse and keyboard actions. OpenAI had introduced function calling for GPT models in June 2023, enabling models to request that software execute specified operations.

In June 2023, the Open Worldwide Application Security Project published its first Top 10 list of security risks for large-language-model applications, including prompt injection and insecure handling of model outputs.

Original sources

4 articles from 4 outlets

Every story links to the reporting it was built from. How we choose sources

Spotted a factual error, a broken link or a misleading summary? Tell us.