Skip to content
Cisco Launches Project Quarry to Detect AI Agent Risks

Cisco Launches Project Quarry to Detect AI Agent Risks

First seen 29 Oct 2026, 03:34 UTC • •
API

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 29, 2026 at 09:38 UTC
  • •Cisco's Project Quarry detects risky AI agent behaviors across full sessions.
  • •The tool includes a command-line scanner and Python SDK for integration.
  • •Quarry aims to prevent data exfiltration by analyzing sequences of actions.

Cisco has open-sourced Project Quarry, a detection engine designed to identify risky behaviors across the full session of AI agents rather than isolated actions. The tool includes a command-line scanner, a Python SDK, and plugins for various AI models. It aims to help security teams flag sequences of actions that may indicate credential theft or data exfiltration, similar to the Hugging Face intrusion. Cisco emphasizes the importance of context in detecting malicious activities, comparing it to credit card fraud detection. The tool is already in use in AI Defense and is set to be integrated into DefenseClaw. The release provides a rule format for describing agent behavior, enhancing the ability to detect complex attack patterns.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated just now How this analysis works

Timeline

2026-10-28
Cisco open sources Project Quarry
Cisco released Project Quarry to help detect risky AI agent behavior across entire sessions, enhancing security measures.
News.Lavx.Hu
2026-10-29
Project Quarry features detailed
The tool includes a command-line scanner and plugins for various AI agents, aimed at improving detection capabilities.
Feeds.4Sysops

More articles in this cluster (4)

Following this threat?

Track Cisco in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Extracted Entities (10)

1 / 3

Common questions

What does Project Quarry do?
Project Quarry detects risky behaviors of AI agents by analyzing their full session activities.
Which AI agents are supported?
The tool supports plugins for Claude Code, Codex, and other MCP-compatible agents.
How can I use Project Quarry?
You can integrate it using the provided command-line scanner and Python SDK to enhance your security measures.