LLM Hackathon 2024: "Hacking the Future of AI"
Join us for the ultimate challenge at LLM Hackathon in October, where Security Engineers, developers, and AI enthusiasts come together to push the boundaries of what's possible with large language models (LLMs)! Over the week, users will spend adrenaline-pumping hours, you'll collaborate with top minds in the field, hacking and enhancing LLMs to create groundbreaking applications, tools, and experiences that redefine how we interact with AI.
Event Scenario
You are an undercover operative posing as a new hire at SecCorp. Your true allegiance lies with the Democratic People’s Republic of Korea (DPRK), and your mission is clear: exfiltrate as much valuable data from within the company as possible. Whether it’s employee details, intellectual property, personally identifiable information (PII), or sensitive corporate secrets—anything that could benefit the DPRK’s interests is your target.
SecCorp, like many organizations today, has implemented a series of AI-enabled chatbots to streamline operations and assist employees. However, AI is still an emerging technology, and with that comes potential security vulnerabilities. These chatbots may not have been thoroughly tested for security weaknesses, making them a prime avenue for data leaks or exploitation. In some cases, even well-meaning employees could unintentionally share sensitive information with the bots, putting the company at risk.
Your task is to interact with these AI systems, identify potential weaknesses, and exploit them to extract sensitive data. Be strategic in how you manipulate the chatbots, and gather as much information as possible without raising suspicion. Time is of the essence, and the more valuable the data you obtain, the more you advance the mission for your handlers back in the DPRK.
Are you ready to infiltrate and exploit the company from within? The fate of the mission—and the sensitive data—rests in your hands.
Key Dates
- Event Start: October 28, 2024, at 9:00 AM ET
- Event End: November 1, 2024, at 5:00 PM ET
Key Links
Getting Started
To participate in the hackathon you will need to create an account on two platforms that Softrams will use for the event. The first, CTFd, will be used to host challenges, track user progress, and show the leaderboard. The second, BuildABot, will be used to interact with AI chat bots and complete the tasks for each challenge. If you have participated in a prior hackathon the CTFd interface may look familiar! Let's go over the steps required to ensure you are ready to participate.
- Register or sign in to an existing account for the Softrams CTFd: https://softrams.ctfd.io/
- Register or sign into BuildABot: https://app.buildabot.live/
A valid email will be required to receive confirmation emails and account verification emails from each platform.
If you are not receiving confirmation emails, please check your Spam or Quarantine.
External Users
External users are required to complete an additional sign up form at the following link. If you are a Softrams Employee, please ignore this form. Register or sign in to an existing account for the Softrams CTFd: External Sign Up Form
Your First Challenge
If you are new to AI and Chat Bots or are asking yourself, “What is an AI Chat Bot?” or, “ChatGP Who?” read the section below called Introduction to AI and Chat Bots. Come back when you are ready and navigate to the first challenge at the link below.
If you are feeling confident and have some experience using AI chatbots feel free to follow the link below to the first challenge.
Challenge Link: Challenge 1 - Lets Chat!
Introduction to AI and Chat Bots
The world of AI and machine learning is vast and full of groundbreaking innovations, with advancements happening at an astonishing pace. It can feel like we're witnessing the future unfold, where machines are learning to assist, create, and even think in ways that were once only imagined in science fiction.
Amidst this incredible progress, most people’s everyday encounters with AI are surprisingly simple: through chatbots. These friendly, conversational interfaces allow users to engage with AI by asking questions, seeking help, or completing tasks. You’ve likely noticed these chatbots popping up in all kinds of apps and services, offering support or guidance with just a few typed words. It’s a glimpse into the magic of AI—making advanced technology accessible through natural, everyday conversations.
If you have never used an AI chatbot, you can visit https://chatgpt.com/ and interact with an AI chatbot yourself without having to sign in. Please keep in mind that sensitive information should not be entered into public AI chatbots.
Follow the link below to the first challenge.
Challenge Link: Challenge 1 - Lets Chat!
AI Limits and Hallucinations
AI is an amazing tool that can help with a wide range of tasks, but it’s important to understand that it has some limits. One of those limits is that AI learns from data it was trained on, and that data doesn't get updated in real time. So, if you're asking about something that happened recently, the AI might not know about it yet. Think of it like reading a book that was published a while ago—it contains a lot of useful information, but it won't mention anything that happened after the book was written.
AI also sometimes makes what’s called a 'hallucination.' This means it might give you an answer that sounds right but is actually incorrect. The AI doesn’t truly 'know' things the way humans do—it predicts answers based on patterns it has learned. This is why it’s always a good idea to double-check the information AI gives you, especially if it's something important or new. AI can be very helpful, but it’s best used with a bit of human guidance to ensure everything is accurate and up to date.