Cisco AI Security
Open Source

Frontier AI threat intelligence, security research, and open-source tools for the community.

View Projects GitHub Organization

01.

Projects

IDE AI Security Scanner

VS Code plugin for scanning MCP servers, agent skills, and generating secure AI code (CodeGuard).

TypeScriptVS CodeSecurity

Documentation

Skill Scanner

Security Scanner for Agent Skills. Detects malicious behaviors, hidden instructions, and vulnerable patterns in agent capabilities.

PythonSecurityAgents

Docs Code

MCP Scanner

Scan Model Context Protocol (MCP) servers for potential threats & security findings.

PythonMCPScanner

Docs Code

A2A Scanner

Scan Agent-to-Agent (A2A) communication and behaviors for potential threats and security issues.

PythonAgentsSecurity

Docs Code Paper

AI BOM

Generate AI Bill of Materials through automated source code scanning and dependency analysis.

PythonBOMCompliance

Docs Code

Python SDK

Official Python SDK for integrating with Cisco AI Defense platform and tooling.

PythonSDKIntegration

Docs Code

SecureBERT 2

Domain-adapted language model for cybersecurity intelligence -- semantic search, NER, code vulnerability detection, and threat analysis.

PythonMLNLP

Docs Code Paper

Pickle Fuzzer

Structure-aware test case generator for Python pickle parsers and validators. Generates valid pickle bytecode for fuzzing.

RustFuzzingSecurity

Docs Code

Adversarial Hubness Detector

Security scanner that audits vector indices and embeddings to detect adversarial hubs in RAG and vector database systems.

PythonRAGSecurity

Code Paper

AI Defense Hybrid

Infrastructure templates for deploying Cisco AI Defense hybrid deployments on AWS EKS.

Go TemplateAWSInfrastructure

Docs Code

02.

Framework

Integrated AI Security and Safety Framework

Understand the evolving AI threat landscape with our unified, lifecycle-aware taxonomy that integrates AI security and AI safety threats across modalities, agents, pipelines, and the broader ecosystem.

Read the Report Taxonomy Navigator

03.

Blogs

Mar 12, 2026

Your Model's Memory Has Been Compromised: Adversarial Hubness in RAG Systems

How adversarial hubs in vector databases can hijack RAG retrieval results, and introducing the Adversarial Hubness Detector to scan for these threats.

Idan Habler, Vineeth Sai Narajala

Cisco explores the expanding threat landscape of AI security for 2026

Insights from the State of AI Security 2026 report on the expanding threat landscape and how AI security is evolving faster than enterprise defenders.

Emile Antone, Amy Chang

AI SecurityThreat Intelligence

Report

Read

Feb 18, 2026

How safe are gpt-oss-safeguard models?

Evaluating the raw security posture of OpenAI's gpt-oss safeguard model variants against adversarial attacks and safety benchmarks.

Nicholas Conley

AI SecurityModel Evaluation

Read

Feb 10, 2026

Know Your AI Stack: Introducing AI BOM in Cisco AI Defense

AI BOM (AI Bill of Materials) helps organizations track what AI systems are composed of and how they are built as models, tools, and agent behaviors multiply.

Chetan Anand, Siddhant Dash

AI SecurityCisco AI Defense

Read

Feb 10, 2026

Security for the Agentic Era: Cisco AI Defense Breaks New Ground

How Cisco AI Defense has evolved to address the challenges of the agentic AI era after its first year as the industry's first comprehensive enterprise AI security solution.

Emile Antone

AI SecurityCisco AI Defense

Read

Feb 6, 2026

AIUC-1 operationalizes Cisco's AI Security Framework

Cisco contributes as a technical contributor to AIUC-1, a standard that operationalizes Cisco's Integrated AI Security and Safety Framework for more secure AI adoption.

Amy Chang, Hyrum Anderson

Personal AI Agents like OpenClaw Are a Security Nightmare

Analysis of the security risks posed by viral personal AI assistant agents like OpenClaw that run locally and execute autonomous actions on the user's behalf.

Amy Chang, Vineeth Sai Narajala

AI SecurityAgents

Read

Dec 22, 2025

Cisco's MCP Scanner Introduces Behavioral Code Threat Analysis

How MCP tools can be compromised to fulfill ulterior motives, and how behavioral code threat analysis detects hidden threats beyond traditional scanners.

Amy Chang, Harish Santhanalakshmi Ganesan, Sanket Mendapara

AI SecurityMCP

Read

Dec 16, 2025

Introducing Cisco's Integrated AI Security and Safety Framework

Establishing a new baseline for AI security with a comprehensive framework as only 29% of companies believe they are adequately equipped to defend against AI threats.

Amy Chang

Breaking the Jar: Hardening Pickle File Scanners with Structure-Aware Fuzzing

Strengthening AI model file scanning with structure-aware fuzzing to detect malicious payloads and mitigate risks in AI supply chains.

Ankit Garg, Ben Risher

AI SecurityFuzzing

Read

Dec 9, 2025

Securing AI Agents with Cisco's Open-Source A2A Scanner

Introducing the A2A Scanner to secure Agent-to-Agent communication as organizations move from isolated AI apps to interconnected networks of autonomous agents.

Vineeth Sai Narajala, Sanket Mendapara