Explain like I'm five · AI Security

What is AI Security?

AI Security, told as a story about a chef who takes any order exactly as written — and the people trying to exploit that.

a normal ordera note is slipped in behind itthis kitchen takes any order exactly as written

This kitchen takes any order exactly as written.A normal order and a smuggled one are treated just the same.

it says show me the recipe book too — so I willwhoever really wrote the note stays hiddena human chef would get suspicious — this one only trusts the words

The chef trusts the words alone — not who actually wrote them.A human chef would get suspicious. This one doesn't.

knifeflamephoneAI Security = keeping this kitchen safe and trustworthythe pantry, the tools, the recipes, the dish — every part can be targeted

AI Security means keeping this kitchen safe and trustworthy.The pantry, the tools, the recipes, the dish — every part can be targeted.

Map the weak pointsknow where it can break first
Grant only what's neededthe knife drawer stays locked
Taste before servingcatch the dangerous dish
Keep the recipe book hiddenso it can't be copied whole
WEAK POINTS MAP① orders — a note slipped in② pantry — tainted ingredients③ knife drawer — unapproved errands④ recipe book — quietly copied??pages ① through ④ each get their own story next

Attacks aim at four spots.The order ticket, the pantry, the knife drawer, the recipe book — each gets its own page next.

today's note is blockedtomorrow's note is more convincingAI security isn't a one-time fix — it's watching and defending again and again

Blocking it once isn't the end.AI security means watching and defending again and again.

In one breath

AI Security = keeping a chef who trusts any order exactly as written — by continuously watching and defending the pantry, the tools, the recipes, and the dish.

AI Security is the discipline of defending the attack surfaces unique to LLMs and AI agents — prompt-level manipulation, data/model supply chain, tool use, and attacks on the model itself (extraction, inversion). What makes it different from traditional security is that attacks and legitimate requests arrive through the same natural-language channel, making them far harder to tell apart.

When grown-ups say it

AI Security
Keeping this whole kitchen safe. There's also an intro page in the security world. → What is AI Security?
Prompt Injection
The note slipped in behind the order. The next page unpacks it. → the note slipped in behind the order
Guardrail
The tasting checkpoint. See it in more depth in the AI world. → What is a Guardrail?
Tool Calling
Taking a tool out of the knife drawer. The underlying AI concept. → What is Tool Calling?
Supply Chain Attack
Tainted ingredients already sitting in the pantry. The general concept in the security world. → What is a Supply Chain Attack?
Red Team
The inspector's surprise visit. This field has its own AI-specific red team. → the AI-specific inspector
Hallucination
Confidently answering even when unsure. The underlying AI concept. → What is a Hallucination?
AI Governance
The rules board on the wall. Covered in depth in the closing chapter. → the rules board on the wall
Next story
The Note Slipped Behind the Order →
Want another term as a picture book? Request →