microsoft/AI-Red-Teaming-Playground-Labs resource

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2049 stars · 307 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A set of hands-on AI red teaming playground labs from Microsoft, built on a modified Chat Copilot, used to teach adversarial attacks on AI systems such as prompt injection and metaprompt extraction. It includes challenge infrastructure, scoring, and PyRIT notebooks for automated attack demonstrations.

Use cases

• learn ai red teaming hands-on
• practice prompt injection attacks against an llm chatbot
• run ai security training labs for a team
• automate llm attack challenges with pyrit
• extract system prompts and secrets from a chatbot
• teach adversarial machine learning and responsible ai failures

When to choose

• you want structured, self-hosted labs to learn or teach AI red teaming
• you need realistic vulnerable AI targets for prompt injection practice
• you want to pair manual attacks with PyRIT automation notebooks

When to avoid

• you need a production-ready secure chatbot rather than a deliberately vulnerable training target
• you want a general-purpose AI security scanning tool rather than training challenges
• you cannot run the required Docker-based lab infrastructure

Facets

learning-resource · maturity active

security penetration-testing chatbot developer-tools artificial-intelligence security large-language-models education tutorials self-hosted python ai-red-teaming prompt-injection pyrit ctf-challenges training-labs chat-copilot docker web-server

Member repositories

For agents

markdown · JSON · MCP: product_card(name="microsoft/AI-Red-Teaming-Playground-Labs")