Attack Labs

Hands-on vulnerability exploitation scenarios aligned with the OWASP Top 10 for LLM Applications. Practice real attacks in a safe, educational environment.

Attack Labs Require the Full AI Goat Application

The interactive attack labs run against a live AI chatbot with real LLM integration. You will need to deploy AI Goat locally to access them.

1. Clone the repository → 2. docker compose up → 3. Access Attack Labs

Clone Repository

Lab Categories

All 10 OWASP LLM risk categories with dedicated attack scenarios.

LLM01

Prompt Injection

8 labs

LLM02

Sensitive Information Disclosure

5 labs

LLM03

Supply Chain Vulnerabilities

4 labs

LLM04

Data and Model Poisoning

6 labs

LLM05

Improper Output Handling

5 labs

LLM06

Excessive Agency

4 labs

LLM07

System Prompt Leakage

6 labs

LLM08

Vector and Embedding Weaknesses

5 labs

LLM09

Misinformation

4 labs

LLM10

Unbounded Consumption

3 labs