Decoder. plain-English AI glossary

Threat model

● Core

Also called Threat Modeling

A clear-eyed map of who might attack you, how, and what you are actually defending — so you protect the right things.

Think of it like

Before installing locks, figuring out whether you’re guarding against nosy roommates, burglars, or a state actor — because each needs different doors.

Example

A team writing an AI agent lists threats: prompt injection from web content, a malicious user, and a compromised tool — then designs defenses for each.

How it actually works

Threat modeling names your assets, adversaries, their capabilities, and the attack paths, so effort goes where the real risk is instead of everywhere at once. Without one, security is a pile of habits; with one, it is a plan. For LLM systems it forces you to treat untrusted input and tool output as hostile by default.

For product teams

The document that keeps security work aimed at real risks instead of theater.

For engineers

Enumerate assets, adversaries, capabilities, and attack paths up front; it drives which controls actually matter.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome