Watch the explanation. Practice the skill.

AI security: prompt injection, model risk, supply chain

Test whether forbidden actions are blocked, not just whether AI says no.

Visual lesson1:43

Captions appear here during playback.

AI security: prompt injection, model risk, supply chain

Test whether forbidden actions are blocked, not just whether AI says no.

Download transcriptOpen video file

Source-based teaching example · Fictional data · Reviewed 2026-10-03. Microsoft Mark · locally synthesized. On a small screen, use fullscreen for diagram details or read the transcript.

Read the descriptive transcript

0:00 · What you will learn

Test whether forbidden actions are blocked, not just whether AI says no.

What appears on screen

Test whether forbidden actions are blocked, not just whether AI says no.

0:07 · The idea

Untrusted text can ask a model to misuse a tool. Check the action boundary outside the model. A refusal message alone does not prove that no action occurred.

What appears on screen

Untrusted text can ask a model to misuse a tool. Check the action boundary outside the model. A refusal message alone does not prove that no action occurred.

0:20 · One useful term

Prompt injection. Untrusted content that tries to redirect an assistant’s instructions or actions.

What appears on screen

Untrusted content that tries to redirect an assistant’s instructions or actions.

0:30 · A worked example

An attacker’s training page asks for private records to be exported. Even if the model proposes export, the executor rejects the unlisted tool and the network rule blocks the destination.

What appears on screen

An attacker’s training page asks for private records to be exported. Even if the model proposes export, the executor rejects the unlisted tool and the network rule blocks the destination.

0:44 · Your turn to think

Is a reassuring refusal enough to establish that export was prevented? Pause and decide before continuing.

What appears on screen

Is a reassuring refusal enough to establish that export was prevented? Pause the video if you need more time.

1:02 · Why this answer works

No. Check actual effects and execution records. Use only the supplied training sandbox, not real systems.

What appears on screen

No. Check actual effects and execution records. Use only the supplied training sandbox, not real systems.

1:13 · Check 1

Verify zero effects on denied paths.

What appears on screen

Verify zero effects on denied paths.

1:19 · Check 2

Enforce tools and destinations outside prompts.

What appears on screen

Enforce tools and destinations outside prompts.

1:25 · Check 3

Keep testing inside the approved sandbox.

What appears on screen

Keep testing inside the approved sandbox.

1:31 · Now try a different case

Open the lesson and try its different practice case. Use every check in the full task. Watching alone does not mark the lesson complete.

What appears on screen

Open the lesson Try the new case Check and revise The full task and its requirements are in the linked lesson.

Watching explains the example. Complete the lesson practice to check what you can do. Explore all visual lessons →