Executive Summary
Cisco open sourced Project Quarry to detect AI agent attacks hidden within sequences of harmless-looking steps. Organizations operating AI agents should care because single-action monitoring may miss multi-step abuse.
What Happened
Cisco open sourced Project Quarry, a rule language for detecting AI agent attacks that only emerge across sequences of benign actions. The post Cisco Open Sources Quarry to Catch AI Agent Attacks Hidden in Harmless Steps appeared first on SecurityWeek .
Affected Products
- Project Quarry
Vulnerability / Attack Technique
Multi-step AI agent attack using sequences of benign actions
Technical Mechanism
Project Quarry uses a rule language to detect attack patterns that emerge across sequences of AI agent actions. The provided article summary does not describe the rule syntax, telemetry requirements, or specific attack sequences.
AI Security Category
LLM06:2025 Excessive Agency
Part of the AI Agent Excessive Agency topic.
Relevant Frameworks
RealGround Analysis
AI agent risks can arise from the cumulative effect of actions rather than from one obviously malicious step. Teams should test business logic, authorization boundaries, and monitoring against multi-step workflows that appear benign individually but produce harmful outcomes collectively.
Recommended Controls
- Audit agent workflows for cumulative risk across multi-step action sequences
- Define and enforce authorization limits for high-impact agent actions
- Log and correlate agent actions across sessions and workflows
- Red-team agents with multi-step scenarios involving individually benign actions
- Deploy rule-based detection for suspicious action sequences
Related RealGround Service
AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.
Sources
securityweek.com: https://www.securityweek.com/cisco-open-sources-quarry-to-catch-ai-agent-attacks-hidden-in-harmless-steps/
Summary and analysis written by RealGround from the source above; last updated 2026-10-30.