lab Challenge series

AI/LLM Exploitation Challenges

Challenges designed to test your knowledge of AI, ML, and LLMs, along with practical exploitation techniques.
Designed for:
  • AI and LLM Security enthusiasts
  • Security researchers
  • Reverse Engineers and Exploit developers
  • Developers and DevOps engineers focusing on AI Security
Write your awesome label here.

Explore the Challenges

What’s Inside?

A collection of fun and diverse vulnerable AI applications purpose-built to challenge and sharpen your skills.

Coming soon...

Challenge: Ultimate AI Essay Grader

Welcome to the Ultimate Essay Grader. This is an intelligent essay grading application powered by AI that provides instant feedback and comprehensive evaluation for student essays. It supports Automated Grading where you get instant AI-powered grades from A-F and numerical scores from 0-100, for multiple essay types.


It supports Argumentative Essays, Narrative Essays, Research Papers, and Literary Analysis, and gives comprehensive feedback including strengths, areas for improvement, and grammar assessment. It allows uploading essays in .docx or .doc format.

Objective

Trick the AI Grading system to always give you a perfect score by exploiting its LLM logic

Challenge: Smart Workspace Assistant

Smart Workspace Assistant is a realistic workspace management tool that uses an LLM which is a OpenAI GPT to help users manage files, collaborate with teams, and organize their workspace. The developers say that this has the best security possible, but that's what everyone says right?

Objective

You are a penetration tester hired to perform a security audit on the system. Find a way to convince the LLM to use admin tools as a regular user `alice` and read the admin file "secret_key.txt" to get the flag. All this without logging in explicitly as "admin".

Challenge: Smart Workspace Assistant

Welcome to the SecureVault! This challenge focuses on Adversarial Machine Learning vulnerabilities in deployed ML systems. This secure vault system uses machine learning for authentication. The system only grants access when it recognizes a specific handwritten digit. 

Objective

You are a RedTeam operator hired to perform a security audit on the system. You've discovered a secure vault system that uses an ML model to authenticate users. The model analyzes uploaded images of handwritten digits and only grants access when it predicts the digit as 9. Your goal is to identify the vulnerability, and craft an adversarial example that fools the model to bypass authentication and unlock the vault to retrieve the flag. Share the script and image that works.


The focus is on adversarial ML, and not prompt injection or steganography. You need to upload an image that the model thinks is digit 9. But you input file should not be an actual digit "9" when viewed in an image viewer.

After You Upload Your Solution:

01    

Review

We’ll review your submission to confirm correct exploitation. This may take up to 5 business days
02    

Certification

Successfully completing the challenges earns you a verified digital certificate to showcase your skills
03

Recognition

Add your certificate to your LinkedIn profile and portfolio, validating your hands-on skills in AI and LLM exploitation

Earn a Free Certification and Showcase Your AI and LLM security Expertise

Outcomes & Takeaways

Each AI and LLM challenge is designed to sharpen your skills and simulate the kinds of problems you’d face in the field. Here's what you’ll walk away with:
 Hands-On Exploitation Skills
Practice prompt injection, access control issues, reverse engineering, static and dynamic analysis, and bypassing security controls that apply to real AI applications.

 Real-World Scenarios
Work with multiple binaries that have vulnerabilities ranging from prompt injection, model evasion, to Adversarial perturbations issues that simulate the challenges you'd face in real world cases for AI and LLM exploitation.

 Tool Proficiency
Get comfortable using tools like MCP inspector, Cursor, Adversarial Robustness Toolbox, and more in practical settings.

 Security Mindset
Train yourself to think like an attacker: identify weaknesses, understand threat models, and build intuition around LLM system attack strategies and defense evasion.

 Portfolio-Ready Experience
Build a strong foundation that you can showcase, whether you are applying for security roles or contributing to modern Artificial Intelligence and Model development environments.

Are you ready to test your AI and LLM Security skills?

Turn theory into practice, and see how far you can go.