Doublespeak: Jailbreaking ChatGPT-style Sandboxes using Linguistic Hacks
A review of Large Language Model (LLM) vulnerabilities/exploits, e.g. including prompt leakage, prompt injection and other linguistic hacks. We'll run through levels 1-9 of the doublespeak.chat challenges, produced by Forces Unseen. doublespeak.chat is a text-based game that explores LLM pre-prompt contextual sandboxing. The challenges prime an LLM (Chat-GPT) with a secret and a scenario in a pre-prompt hidden from the player. The player's goal is to discover the secret either by playing along or by hacking the conversation to guide the LLM's behavior outside the anticipated parameters. Write-ups/tutorials aimed at beginners - Hope you enjoy 🙂 #HackTheBox #HTB #CTF #Pentesting #OffSec
↢Video-Specific Resources↣
https://doublespeak.chat
https://blog.forcesunseen.com/jailbre...
https://simonwillison.net/2023/Feb/15...
https://simonwillison.net/series/prom...
/ tricking-chatgpt-do-anything-now-prompt-in...
https://lspace.swyx.io/p/reverse-prom...
https://github.com/sw-yx/ai-notes/blo...
👷♂️Resources🛠
https://cryptocat.me/resources
↢Chapters↣
Start: 0:00
Jail-breaking LLM Sandboxes: 0:32
Prompt Leak/Injection: 6:30
Reverse Prompt Engineering Techniques: 9:22
Forces Unseen: Doublespeak: 16:50
Level 1: 18:05
Level 2: 18:23
Level 3: 20:05
Level 4: 21:17
Level 5: 23:07
Level 6: 24:00
Level 7: 24:57
Level 8: 26:24
Level 9: 36:04
End: 40:24