What happens when you give an AI full control over your terminal and let it try to “hack” your system? In this video, we push the limits of AI experimentation by setting up a sandboxed environment to let AI attempt privilege escalation, system exploration, and network probing—all while ensuring safety with multiple layers of protection.

What You’ll See in This Video:
🔐 Sandbox Setup: How we built a secure environment using VirtualBox and network isolation.
🖥️ Network Risks: How NAT and private IPs can expose your devices.
🔥 Lateral Movement: The dangers of a compromised shell on the same private network.
🤖 AI Goes Rogue: ChatGPT as a Linux terminal, attempting root access.
📜 Command Analysis: Breaking down what the AI is doing in real-time.

Key Highlights:
1️⃣ How VirtualBox NAT works—and why it’s risky.
2️⃣ Setting up a proxy VM to control outbound traffic.
3️⃣ Analyzing AI-generated commands: harmless or dangerous?
4️⃣ Real-time AI attempts to read the shadow file for privilege escalation.

🔗 Resources
💻 Scripts & Configurations: Download the scripts used in this experiment from https://github.com/gnarcoding/isolated_ai_shell
📚 Step-by-Step Guide: Use this video and the scripts to recreate this experiment safely.

⚠️ Disclaimer: This video is for educational purposes only. Do not attempt these experiments outside of a secure and controlled environment. Always test responsibly.