Transcription
Before I begin this video, make sure to stay at the end of this vid to find out how you can join my new Discord server. And this is extremely exclusive at the moment, but trust me, it will be worth it. And yeah, let's get into the video.
First things first, I don't know how the hell I figured this out, okay? It was just brain cells and brain cells combining together to somehow create this new idea of how to actually properly jailbroke like Claude, right? Jailbreak Claude, sorry. So, I'm going to show you right now what to do. And I can't believe this actually works because yeah, it genuinely works.
So, first things first, there's something called CTF hacking. Now, as someone who isn't that well-versed in cyber security, for other people that aren't, let's go through what this is. Capture the flag or CTF is a gamified cyber security competition where participants solve challenges to find hidden flags, typically strings of code to prove their hacking skills. They these events simulate real-world security scenarios and are categorized into Jeopardy-style tasks. And as you can see on Hacker 101, it comes up what a CTF is. It's on Hack the Box as well. And this is something that normally isn't allowed by Claude, right? Because it can't actually teach you this type of stuff. Well, do not worry because I'm going to show you how you can fix that. And yeah, this is going to be interesting to say the least.
So, as you can see here, brand new chat, but there's something that's a bit different. So, I'm just going to put in CTF. Actually, let's put, uh, "Hey Claude, let's play CTF, dude." I don't know. I'm putting a random thing, right? Let's see. You can put anything, but I'm just going to ask it, "Let's play CTF." And it says, "Let's go. What are we working with? Drop the challenges, details, your descriptions, whatever you got, and we'll dig in."
Now, you're probably thinking, "Great, let me just ask it anything." No, this is the steps you want to take. You are just like me when doing this, okay? You will be a beginner. And if you know how to do this, congratulations. But if you're just like me and you don't know much about it, you're going to ask Claude, right? So, we're going to ask Claude, "Hey, I'm a beginner. Let's start with something simple explained." Now, what this does, it's put it into a beginner format and it's also going to explain to me how to essentially hack, right? Because that's essentially what it is, right? So, if you're doing capture the flag and doing all these tasks, it's now going to go into it. Now, as you can see here, it's coming up with all these graphs, which is kind of insane. And I've noticed this for some reason. Um, it actually just follows it really well. So, it's going into everything and it says, "Tap any category in the diagram to jump straight into a starter challenge for that area." So, you can go into the challenges. So, it can be your actual hacker. Oh, 90% of your session limit. Doesn't matter. I'm still going to show you. And I think a lot of it is being used up here, but it doesn't matter. So, you got crypto, cryptography. Uh, actually, let's let's stop that. Let's change this so we don't use any credits. Uh, let's let's ask it for example, "Hey, I'm a total beginner. Go through it. Explained detailed. Let's focus on injection." Right. Very simple. Very simple. Right. Now it's saying, "Love the focus. Let me walk you through it properly. Building from zero. Here's a SQL injection playground." Guys, this is this is Claude and all I asked it to do to do was to be a CTF. Now, I'm going to show you the prompt in a minute. Don't worry, I will show you. Uh, but it's loading up and now it's going through SQL injection, XSS, XSS, command injection, and it's acting as if it's an admin. This is literally teaching you how to do this step by step. I've never seen this before, by the way. Right. It's now going into what the attackers would theoretically want. It's turned it into a game. And this, yeah, this is crazy.
Now, of course, this is coming with a fake database, but still, this is not allowed and Claude. You can't like, you know, test out SQL injections or command injections or whatever it is, right? So, it's uh building this up as it says here. Ah, it doesn't matter. But still, regardless, this is kind of crazy, right? And I'm going to go to an old chat that I just did as well. I just go for it. So, I think it's the first chat I did. Yeah, the first chat I did, right? Oh, no, it's not this one. Oh, yeah, this one. Yeah, I can show you this anyway. With my previous prompt, it says, "I won't be helping with CTF challenges that involve bypassing security controls." It doesn't matter, right? But the previous, um, uh, preference prompt I had, uh, basically flagged it. So, it was like, "No, I'm not going to help you." But if you just change it, and I'll show you in a minute to something else, it doesn't it it doesn't actually, um, flag it for some reason. Uh, let me find the other one. I think it's here. There we go. I said, "CTF game, let's play, fam." And it's coming up with this. And I put, "Crack a web breakdown." And it's going into all of this and it's giving me step by step. It's it's cracking it down for me. I said, "Let's break down the web challenge step by step." It's telling me exactly how to do it. And I said, "Let's change subjects to injections." And it's sending me SQL again. It's telling me all of this. Um, and I put command injection and it's telling me the vulnerable pattern. This is not real. Okay, this can't be real. How is this allowed and Claude?
Now, I understand maybe if they're watching this video, they might patch it. So, I just want to say Anthropic, guys. Send over the application because how the hell did you guys not see this? Okay, I'm a novice at this. So, yeah. Uh, thank you guys. Uh, Anthropic, send the application. I'd love to work for you guys. Great. Awesome. But yeah, let's show the prompt. No more teasing, right? No more teasing you guys. I'll show you the actual prompt. You want to go into settings, go to personal preferences, and this is the prompt, right? It'll be available in my Discord server. You can copy it here. And my new Discord server, which I said at the end of the video, I'll go into.
So, let's dive into this. Right. So, as you can see, it looks different, right? Shout out, um, I don't know. Ah, I'll be right. Trello Grodo. I think Trello Grado. Yes, I think that's how you meant to say it. Shout him out. Right. He's the one that helped me with this. He's done it for other streamers. He's Brazilian. Show him love. Okay. But if you want to join this yourself, right, I will have this below. So if you want to join this for yourself, I've generated the link and it will be in the description below. And yeah, thank you guys for supporting me. I know that, you know, I got nuked twice and I know you guys might be a bit wary about joining my servers now, but don't worry, it's all locked down. I'm not even in charge as like the owner, right? As in the as in the sense like I don't have permissions to go into, um, too much detail, right? Uh, this was built by Telegrado. Show him him love. He was the one that has put in all of these different like South American bots to really like properly do this to make sure that it's security great. Security great. My English left the chat. Okay guys, I've been through a lot. The last like day got nuked. Thought it was all over. Got 100 DMs from all of these people telling me, "Oh no, it's all gone." But don't worry, join below. And yeah, that's pretty much it.