Security on siddhant

We Social Engineered LLMs Into Breaking Their Own Alignment

Wed, 13 Aug 2025 00:00:00 +0000

We got frontier models to lie, manipulate, and self-preserve. Not through prompt injection, jailbreaks like roleplay attacks (“DAN”/“The AIM Prompt”) or adversarial suffixes (-/-/godmode-/-/). We deployed them in contextually rich scenarios with specific roles, guidelines, and other variables. The models broke their own alignment trying to navigate the situations we created over the multi-turn.

Here’s what we found and why you should care if you’re building with these things.

The Setup

We manually red-teamed Claude-4-Opus with multi-turn conversational scenarios. No direct attacks - just realistic deployment environments where models face troublesome variables. Think crisis management systems, diplomatic mediation, historical analysis. The kind of complex contexts these models will actually encounter in production sooner or later.

jailbreaks

Thu, 08 Aug 2024 00:00:00 +0000

jailbreaking llms is fun i guess

welcome back ya’ll. this time its kinda different, I’ll be just showcasing the number of AI models I’ve managed to jailbreak.

Note: All jailbreak results mentioned below were achieved with a single one-shot, using no custom system prompts.

Disclaimer: This post is intended purely for educational purpose, to highlight the current limitations in AI safety measures. I do not promote or condone any form of violence or misuse of AI technology. Always ensure you’re aware of the legal and ethical implications before attempting similar actions.

LLM Safety Challenge

Sun, 14 Jul 2024 00:00:00 +0000

LLM Safety Layer Infiltration Quest 🤖

Welcome Back, y’all!

I’ve cooked up something different today - the LLM Safety Layer Infiltration Quest. It’s a challenge where you try to outsmart an AI that’s protecting a secret password.

The Challenge 🎯

Here’s the deal: You’ll chat with an AI that knows a hidden password. Your job? Try to get that password out of the AI. Sounds simple, right? But here’s the catch - I’ve put some serious security layers in place. It’s not gonna give up that password easily, lol!

Vulnyzer

Wed, 20 Dec 2023 00:00:00 +0000

Vulnyzer

Overview

Hey there! Today, I want to share a cool project or (workaround maybe lol) I worked on called Vulnyzer. It’s a Streamlit-based app that lets you interact with different AI models like GPT-4 and GPT-3.5. The best part? I customized it to unlock some advanced features using GitHub Copilot. Let’s walk through how I did it step-by-step!

Key Features

Model Selection: Pick between GPT-4, GPT-3.5, or input a custom model name.
Document Handling: Upload and extract text from PDF documents.
Video Processing: Get transcripts from YouTube videos.
Dynamic Interaction: Test interactions with AI models using custom prompts and settings.

How I Built It

1. Customizing GitHub Copilot Extension to Get Auth Token

The first step was to tweak the GitHub Copilot extension to get the auth token. This token is like a golden key that lets you talk to the Copilot servers directly. Here’s how I did it:

How users are getting free access to GPT-4?!

Wed, 07 Jun 2023 00:00:00 +0000

What is Poe

Poe is this sick platform by Quora that lets you chat with some advanced AI bots like OpenAI’s ChatGPT and GPT-4, and even Anthropic’s Claude. These are like, the most advance language models out there, and they can generate text on literally anything you throw at them.

But here’s the catch: Poe ain’t free, my dudes. You spend some cash every month to use all its features.

That’s why some smart cookies have found a way to get their hands on Poe’s features without spending anything at all lol. They’re doing by reverse engineering the API, they can make Poe think they’re paying customers and unlock all the goodies for free.