0din-ai/ai-scanner

AI model safety scanner built on NVIDIA garak

What it solves

It provides a way for organizations to perform security assessments on AI models before they are deployed. It acts as a penetration testing tool specifically for AI systems, identifying vulnerabilities like jailbreaks and other security flaws that could be compromised in a production environment.

How it works

Built with Ruby on Rails and powered by NVIDIA garak, the application runs a series of "probes" (attack prompts) against target AI systems. These targets can be API-based LLMs or browser-based chat interfaces. The system tracks the Attack Success Rate (ASR) and provides detailed logs and reports on which probes succeeded in bypassing the model's safety filters.

Who it’s for

Security teams, AI developers, and organizations deploying LLMs who need to verify the safety and robustness of their AI systems against known attack vectors.

Highlights

  • Extensive Probe Library: Includes 179 community probes across 35 vulnerability families, aligned with the OWASP LLM Top 10.
  • Real-world Attacks: Ships with actual 0DIN-disclosed jailbreaks and multiple retargetable variants of each attack.
  • Enterprise Features: Supports multi-tenant deployments, scheduled scans, and SIEM integration (Splunk or Rsyslog).
  • Comprehensive Reporting: Offers PDF exports with drill-down capabilities and trend tracking for Attack Success Rates.

Related

  • Project
    openai/codex-security

    A CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities in code using AI, featuring support for multiple LLM providers and bulk scanning.

  • Project
    asamassekou10/ship-safe

    An AI security scanner that detects vulnerabilities in application code, AI agents, and supply chains, specifically targeting risks like prompt injection and agent hijacking.

  • Project
    msoedov/agentic_security

    An open-source vulnerability scanner for LLMs and agent workflows that protects AI systems from jailbreaks, fuzzing, and multimodal attacks.

  • Project
    confident-ai/deepteam

    An open-source red teaming framework for LLM systems that simulates adversarial attacks to uncover vulnerabilities and provides guardrails for production security.

  • Project
    ASCIT31/Dark-Moon

    An open-source AI penetration testing platform that autonomously orchestrates specialized agents to run full security assessments across web, cloud, and network infrastructure.