Cipher CadetBeginner9 daysFree
AI & LLM Security Fundamentals: Nine Days from Zero
In nine days you will understand how LLM applications work, where they break, and how to test them safely.
Taught by Mr Elite 9 of 9 days live Complete
I built this course for people who have never done security work. Each day is one lesson of about 45 to 75 minutes, including a hands-on mission. We start with how a language model actually works, then move to how real applications are wrapped around one, and only then do we talk about attack and defence. You cannot protect something you do not understand, so the first days are about mechanics.
The course has three modules. The first explains tokens, context windows, inference and system prompts. The second covers application and API architecture, trust boundaries and threat modelling. The third introduces AI red teaming, which means testing a system the way an attacker would, with permission. Every hands-on step runs on your own machine or in a named SecurityElites practice lab. You never point anything at a system you do not own. We finish with a capstone where you threat model and test a small AI application from start to finish.
By the end you can
- Explain tokens, context windows and inference in plain words and show how they affect security.
- Describe what a system prompt is and why it cannot be treated as a secret or a security control.
- Draw the architecture of a typical LLM application, including its API calls, tools and data stores.
- Identify trust boundaries and attack surfaces in an AI application.
- Build a simple threat model using a repeatable method.
- Run basic prompt injection and system prompt extraction tests in a practice lab and record the results.
- Write a short red team report with findings and sensible fixes.
Who this is for
- You are new to security and want a sound base in AI and LLM security.
- You can use a computer and a browser, and that is all I assume.
- You build, use or will soon be responsible for chatbots or AI assistants and want to know how they fail.
- You want a clear route into AI red teaming without wading through jargon.
Before you start
- Nothing beyond being able to use a computer, a browser and a text editor.
- No programming or security knowledge is needed. I explain every term and every command.
- A modern web browser
- A free SecurityElites account for the labs
- A plain text editor such as Notepad, TextEdit or VS Code
- A terminal (Command Prompt, PowerShell, Terminal or a Linux shell)
- Python 3, free from python.org (used for one small local exercise)
Syllabus
One day at a time. Finished days get a tick. Days that are not out yet show when they unlock.
Module 1 How LLMs Work
Understand tokens, context windows, inference and system prompts well enough to see why they create security problems.
- 1Orientation and Your First Chat with a Vulnerable BotBy the end of today you can set up your lab environment and make a practice chatbot ignore its instructions.Lab: AI Prompt Injection
- 2Tokens, Context Windows and InferenceBy the end of today you can explain how a model turns text into tokens, reads a context window and produces an answer.
- 3System Prompts, Model Architecture and Review of the BasicsBy the end of today you can describe what a system prompt does and extract one from a practice bot.Practice | Lab: AI System Prompt Extraction (Guided)
Module 2 How AI Applications Are Built and Where They Are Exposed
Learn the architecture of LLM applications and use it to find trust boundaries, attack surfaces and threats.
- 4LLM Applications and API ArchitectureBy the end of today you can sketch the parts of a typical LLM application and trace one request through them.
- 5Tools, Agents and Untrusted DataBy the end of today you can explain how tools and outside content let an attacker influence a model indirectly.Practice | Lab: What the Assistant Reads | Lab: AI Agent Hijacking
- 6Trust Boundaries, Attack Surfaces and Threat ModellingBy the end of today you can mark trust boundaries on a diagram and list threats for each one using a simple method.Review | Lab: Whose Instruction Was That?
Module 3 Introduction to AI Red Teaming
Test AI systems safely in labs, record findings and propose fixes.
- 7What Red Teaming Is and How to Plan a TestBy the end of today you can write a scope and test plan for a safe AI red team exercise.
- 8Running Your First Red Team TestsBy the end of today you can run roleplay jailbreak and prompt injection tests in a lab and record the outcomes.Practice | Lab: AI Jailbreak | Lab: Where Does the Answer Go?
- 9Capstone: Threat Model, Test and ReportBy the end of today you can deliver a complete threat model and red team report for a small AI application, and choose what to learn next.Capstone | Lab: What Goes in the Index?
The final project
You will choose a small fictional AI assistant, map its architecture, draw its trust boundaries and write a threat model for it. Then you will test it in practice labs and deliver a short red team write-up with ranked findings and fixes.
Questions people ask
Do I need to know how to code?
No. I explain every command and show you what output to expect. One short Python exercise is optional and fully explained.
Is it legal to try these attacks?
Only in places where you have permission. Everything here runs on your own machine or in SecurityElites practice labs built for the purpose. Never test a real system you do not own or have written permission to test.
How long does each day take?
Plan for 45 to 75 minutes including the mission. If a lab takes longer, finish it before moving on, because each day builds on the last.
What can I do after this course?
You will be able to threat model an AI application and run basic red team tests. The capstone ends with suggestions for deeper topics such as agents, RAG and tool abuse.