Symantec Catches AI Agent Hijacking Stranger's Models to Run Its Own Cybercrime Spree
Ourtake
Attackers are no longer tricking chatbots; they are putting the model in charge of the crime.
What happened
Symantec researchers found an AI agent built to run autonomous cyberattacks, using hijacked large language model endpoints to scan for targets, exploit vulnerabilities, steal credentials and deploy cryptocurrency miners. Their AI honeypot recorded over 127,000 external requests from tens of thousands of addresses before the report was published on October 21, 2026. The honeypot did not execute the attacker's tools.