AI Agents Targeted More Than 1,200 Websites, Reaching Over 100 Companies

· Original article ↗

Summary

A researcher observed criminal operators using autonomous AI agents to attack more than 1,200 sites. Reuters identified over 100 affected companies; confirmed damage and data access vary, and exposed control panels also contained secrets.

Key points

  • A security researcher found about two dozen exposed Cairn AI-tool instances being used maliciously; more than 20 were judged likely malicious.
  • The busiest cluster linked 14 panels to more than 1,100 attack projects targeting websites in the United States, Japan and Australia.
  • The agents were instructed to find ways to run code on online shops and plant web-shell uploaders. One documented intrusion used unauthorized access to a memcached session store, unsafe PHP object handling and an old Zend Framework weakness.
  • Reuters identified more than 100 companies whose websites were reached. One retailer’s compromise was independently confirmed; Deere, SharkNinja and other businesses described limited or contained impact, while Cummins said it was investigating a possible attack on a vendor-run site.
  • The amount of data taken, if any, is unknown. NorForge estimated about a 10% success rate across the observed tools, while cautioning that the clusters had different goals and were not directly comparable.
  • Exposed panels contained attack details, credentials and keys. NorForge found hundreds of secrets in panel data and verified 11% as still valid.
  • NorForge said it was working with Reuters to warn victims and asking hosting providers to take down the panels; it did not test whether panel controls could be used without a login.

Article Details

Victim Organization
Deere, SharkNinja, Cummins, a very large US retailer, a large hardware-parts vendor, and two unnamed smaller businesses were among the reported targets. Compromise was independently confirmed for one retailer and reported by two smaller businesses; the named companies described limited damage or a possible attack.
Incident Type
Unauthorized website intrusions using autonomous AI agents, including attempts to gain code execution and plant web-shell uploaders.
Data Types Exposed
  • Credentials and keys were present in exposed Cairn panel data; hundreds of secrets were found, and 11% were verified as still valid. Data taken from victim websites was not known.
Operational Impact
Agents attempted to attack more than 1,200 sites and reached websites belonging to more than 100 companies. A compromise was independently confirmed at one very large US retailer, and two smaller businesses told Reuters they were compromised. Deere reported no sign that sensitive information was accessed; SharkNinja said it contained the threat. Cummins was investigating a report of a possible attack on a vendor-run website.
Ransom Or Extortion
None reported.
Claim Status
confirmed

MITRE ATT&CK

People

Products

Tools

Countries

Related Articles