Cisco Launches Project Quarry to Detect AI Agent Threats

Cisco has released Project Quarry, an open-source detection tool designed to identify potentially dangerous AI agent behavior by analyzing sequences of actions rather than isolated commands. The platform combines machine learning, large language model analysis and rule-based checks to help security teams detect threats, investigate activity and respond to suspicious behavior.

Cisco has released Project Quarry, an open-source security tool designed to detect potentially dangerous AI agent behavior by analyzing the sequence of actions an agent performs. The technology addresses a limitation in conventional security controls: individual operations may appear legitimate even when their combined effect exposes credentials, bypasses restrictions or transfers sensitive data to unauthorized destinations.

The tool introduces a detection language that allows security teams to define patterns of suspicious behavior, evaluate agent activity and examine the evidence behind each finding. Rather…