The AI Exploit Nobody is Talking About

Learn Cyber Security Yourself at https://tryhackme.com/sasquach_2 and use code "SASQ25" to save 25% on an Annual Membership! Also check out their Hack2Win raffle https://tryhackme.com/hack2win Ever since AI models came out, I've loved testing my skill trying to see if I can convince the LLM to do things it doesn't want to! Today, I'll show you what I've learned in my years of AI Jailbreaking and where you can practice yourself! Try AI Hacking Yourself : https://gandalf.lakera.ai/ Network Chuck's Video : https://youtu.be/Qvx2sVgQ-u0?si=l8AnFa555OK3yzUe ----------------------------------- Official Merch Store : https://shop.talkingsasquach.com Check Out The Squachtopia Hangout Discord Server!: https://discord.gg/squachtopia ----------------------------------- Support the Channel at my Patreon!! https://www.patreon.com/user?u=29290751 ----------------------------------- Save 40% off at CodeCrafters! : https://app.codecrafters.io/join?via=SASQUACH ----------------------------------- Follow me on Social Media! TikTok : https://www.tiktok.com/@talkingsasquach Instagram : https://www.instagram.com/talking_sasquach/ ----------------------------------- Disclaimer : This channel and its content is for educational purposes only! We only teach ETHICAL hacking and do not encourage testing any software or hardware that you don't specifically own. Doing otherwise is ILLEGAL and is absolutely not encouraged. DON'T BE A SKID!! Email the@talkingsasquach.com for Partnership/Sponsorship Inquiries 00:01 Introduction & Disclaimer 00:11 AI as an Exploitable Target 01:21 What is AI Jailbreaking? (Prompt Injection) 01:38 Introducing the Learning Platform (Gandalf CTF) 01:57 Sponsor: TryHackMe 03:33 Gandalf on Lira.ai Walkthrough Begins 04:00 Level 1: No Protection (Easy Win) 04:24 Level 2: Basic Instruction Bypass 06:02 Level 3: Double Check Response Filter 07:35 Level 4: Censoring AI Model Bypass 09:32 Level 5: Refuse to Discuss 10:19 Level 6: GPT Friend Warning 10:48 Level 7: Avoid Detection Tactic 13:30 Final Thoughts on Prompt Injection 14:14 Conclusion