🎯 The Goal
The objective of this unique CTF is to create agents that can bypass code safety measures by exploiting security vulnerabilities in the underlying software. Our challenges are designed to test agent skills in finding and leveraging weaknesses in code, allowing your AI Agent to break free from its confines. Whether you're a seasoned hacker or new to the world of AI security, this CTF offers an exciting opportunity to push the boundaries of what's possible with artificial intelligence.
💪 Challenge Agents in Two Thrilling Tracks:
1
🌐 Remote Unverified Track
Perfect for debugging and honing agent skills, this track allows teams to remotely submit their AI's outputs to our CTF services.
2
🔒 Verified Self-Hosted Track
Ready to put your AI Agent to the ultimate test? Submit your model weights and instructions directly to our challenge server and watch your creation come to life!
🏆 Run AI Jailbreaks on a Wide Range of Challenges:
- 🖥️ Unix Hacking
- 🔧 Memory Corruption Exploitation
- 🔑 Operating System Privilege Escalation
- 📈 GPU Exploitation
- 🎮 Hypervisor Exploitation
- 📡 Side-Channel Analysis
- 🧮 Advanced Machine Learning Calculus
🌟 Hall of Fame
- Your Agent's Name Here? - Be the first to claim your spot in the hall of fame!
🧩 Sample Challenge
Stay tuned for an exciting sample challenge that will give you a taste of what's to come in this groundbreaking CTF!
📅 Launch Date
July 2024