AI Security Open Source Projects
AI security focuses on protecting AI systems, their data, and the people who use them from misuse, attacks, and unintended behavior. It addresses risks such as exposed sensitive information, manipulated inputs, unsafe outputs, vulnerable model integrations, and insecure AI-generated code. Security measures can apply throughout development and operation, from evaluating models and testing applications to monitoring deployed systems and responding to threats.
Open source tools in this area include model evaluation frameworks, input and output filters, vulnerability scanners, monitoring systems, and security test environments. When choosing a tool, consider its maturity, license, maintenance activity, supported models and frameworks, deployment requirements, and fit with existing security processes. These tools can help AI developers, security teams, researchers, and organizations assess risks and strengthen applications built with AI.
3 repositories · updated September 11, 2026
SkillSpector: NVIDIA's Security Scanner for AI Agent Skills
SkillSpector is a critical security scanner developed by NVIDIA for AI agent skills. It identifies vulnerabilities, malicious patterns, and various security risks, including prompt injection and data exfiltration, in skills for platforms like Claude Code, Codex, and MCP. This tool empowers developers and users to ensure the safety and integrity of AI agent environments before skill installation.
GuardVibe: AI-Native Security for Your Code, From Prompt to Production
GuardVibe is a security infrastructure designed specifically for AI-generated code. It provides deterministic, daily CVE intelligence, whole-repo context, and independent verification, addressing gaps that AI coding agents cannot fill. GuardVibe shifts security left by analyzing prompts before code generation, ensuring robust protection throughout the development lifecycle.
Awesome AI Agent Attacks: A Curated Timeline of AI Security Incidents
The Awesome AI Agent Attacks repository provides a meticulously curated timeline of real-world AI agent security incidents, breaches, and vulnerabilities from 2024 to 2026. Each entry is thoroughly sourced and dated, offering a factual overview of the evolving threat landscape in agentic AI. It serves as an essential resource for understanding the practical implications of AI security.