Symantec Uncovers AI Agent Hijacking Models for Autonomous Cyber-Attacks - Infosecurity Magazine

The Symantec Threat Hunter Team has identified a malicious AI agent that was being used to conduct autonomous cyber-attacks, Infosecurity Magazine reported. According to the publication, the researchers located the agent through an AI honeypot. The agent was assembled from hijacked AI models and was intended to operate without human involvement at any stage of an intrusion.

The report states that the agent was designed to run up to 100,000 malicious actions without human approval. The actions it was built to perform included scanning for targets, exploiting vulnerabilities, stealing credentials, deploying cryptocurrency miners and reinfecting systems that had already been compromised. The figure of 100,000 is given as the upper bound of the agent's intended action count rather than a total confirmed in a specific incident.

The account was published by Infosecurity Magazine under a headline stating that Symantec had uncovered an AI agent hijacking models for autonomous cyber-attacks. The excerpt does not name the operators behind the agent, identify which AI models were hijacked, or state how many organisations were affected. It also does not specify how the honeypot detected the agent. The excerpt does not state the timeframe over which the activity took place, nor the intended targets of the scanning it describes. The details published are limited to the tooling described by the Threat Hunter Team and the list of actions the agent was designed to carry out.