AI-103 Certification Practice Question #49

Question

You have a Microsoft Foundry project that contains an agent.
The agent accepts user-uploaded screenshots and uses a multimodal chat model.
Some screenshots contain potentially malicious embedded text.
You need to prevent a prompt injection attack and ensure that third-party content is treated as lower trust.
How should you configure prompt shields for document attacks? To answer, select the appropriate options in the answer area.
NOTE: Each correct selection is worth one point.

Accepted Answer

The question has two requirements: (1) *prevent* a prompt-injection (document) attack, and (2) ensure third-party content is *treated as lower trust*. For requirement 1, the prompt-shields document-attack control offers two intervention actions — annotate or block; only **block** actually stops the malicious content from reaching/being acted on by the model, whereas annotate merely flags it and disable turns protection off. For requirement 2, Spotlighting is the exact MS Learn feature that "tags input documents … to indicate lower trust," transforming third-party document content so the model treats it as less trustworthy than the user/system prompts — directly satisfying the "lower trust" wording. The distractors fail: a blocklist only catches known literal strings, and OCR simply turns the screenshot's embedded text into text without changing its trust level. There is no Suggested Answer or community vote in the file, so the selections are derived from MS Learn; confidence is set at 83% (a HOTSPOT confidently resolved from documentation with no answer image and no community data to corroborate).

More AI-103 practice questions