microsoft/RAMPART

A pytest-native safety and security testing framework for agentic AI applications

What it solves

RAMPART addresses the challenge of testing the safety and security of AI agents. It provides a structured framework to identify adversarial attacks, benign failures, and various harm categories, ensuring that agentic AI applications are robust and secure before deployment.

How it works

Built as a pytest-native framework, RAMPART allows developers to write safety and security tests using a familiar Python testing tool. It uses evaluation-driven assertions to measure risk and assess whether an agent's behavior meets safety requirements.

Who it’s for

This tool is designed for developers and security researchers who are building and deploying agentic AI applications and need a rigorous way to perform red teaming and risk assessment.

Highlights

  • Pytest-native integration for a developer-friendly experience.
  • Support for testing adversarial attacks and benign failures.
  • Coverage of a broad range of harm categories.
  • Evaluation-driven assertions for objective risk measurement.

Related

  • 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
    Giskard-AI/giskard-oss

    An open-source Python library for testing and evaluating agentic systems, providing tools for automated red-teaming, vulnerability scanning, and LLM-as-judge evaluations.

  • Project
    AISecurityLab/hackagent

    A security red-teaming toolkit for AI agents that automates the discovery of vulnerabilities like prompt injection and jailbreaking using research-backed attack techniques.

  • Project
    tophant-ai/aibeat

    A security evaluation framework for generative AI that provides evidence-based safety testing for LLMs, RAG apps, and AI agents.

  • Dispatch
    Anthropic outlines challenges and best practices for red teaming AI systems

    Anthropic released a detailed analysis of red‑team­ing methods—expert, automated, multimodal, and crowdsourced—highlighting their benefits, challenges, and policy recommendations to standardise AI safety testing.