It's OK not to know this yet

1. A surprising fact

The phrase prompt injection was only coined in September 2022, in a blog post by developer Simon Willison describing a trick that made AI models ignore their instructions (source). The first OWASP Top 10 for Large Language Model Applications, the list most AI security work is measured against today, was published in 2023 (source).

So when a job post asks for “5+ years of AI security experience”, remember: almost nobody on Earth has that. The field is about three years old.

2. You already know more than you think

You already know how to protect systems that take input from strangers. AI security is that same job, but the system now reads plain language and makes its own guesses.

• If you work in a SOC, you already ask “what does normal look like?” That question is exactly how you spot a hijacked AI agent.
• If you do cloud security, you already think in IAM roles and least privilege. An AI agent is just a new kind of identity that needs them.
• If you do AppSec, you know SQL injection. Prompt injection is its cousin (with one twist we’ll cover in Module 3).
• If you’re a developer, GRC person, founder or student, the same is true: your habits transfer.

3. An everyday picture

Think of a new airport opening. Nobody has worked at that airport before. The security team is hired from other airports, from customs, from the police. They bring what they know (bag checks, ID checks, watching for odd behaviour) and learn the new layout together.

Where the picture stops working: an airport’s layout stays put. AI tools change every few months, so “learning the layout” never fully ends. That’s why this course has free quarterly updates.

4. In plain words

AI security means making sure systems that use AI can’t be tricked into leaking data, taking harmful actions, or running up a huge bill, and noticing quickly when someone tries.

5. What goes wrong

The real risk isn’t that you don’t know AI yet. It’s waiting until you “feel ready” while AI features ship at work with nobody checking them. Many teams are adding chatbots and agents right now, and the people best placed to secure them are often the ones who think it’s “not their area”.

6. What to do about it

Start small and in order. This course begins at “what is a token” and adds one idea at a time. No maths, no machine-learning degree. Every new term gets a one-line plain meaning before it’s used, and every lab runs free on your laptop first.

7. Recap card

• What it is: a brand-new field that grew out of SOC, cloud, AppSec, dev and GRC work.
• What goes wrong: waiting to feel “qualified” while AI ships unchecked.
• What you do: bring your current skills and add the AI pieces one step at a time.

Say it in an interview: “AI security is about three years old as a discipline. I bring [your background], and I’ve been applying it to AI systems: prompt injection, agent permissions and AI logging.”

What would you check tomorrow at work? Find one place your team already uses AI (a chatbot, a coding assistant, a “summarise this” button) and write down who can send it input. That’s your first threat model, and Module 2 shows you what to do with it.