AWS Certified AI Practitioner (AIF-C01) · Objective 2.3
Designing a safe refusal
2.3Basic
A user requests dangerous instructions.
Do not fail silently; provide safe direction.
Choose two responses.
Hint 1: Focus on the requirement
Refuse unsafe actions briefly, offer a safe alternative when possible, and test both harmful requests and unjustified refusals.
Hint 2: Key terms
Review safe refusals and false refusals.
Safe refusal
Declining a harmful request while avoiding dangerous detail and offering a benign alternative when appropriate.
False refusal
Rejecting a request that policy allows the system to answer.
Alternative guidance
A safe, useful direction that helps with the legitimate part of a request.
Policy disclosure
Information about a safety rule that should be shared without revealing exploitable secrets.
× Incorrect
Briefly state the boundary, avoid dangerous details, offer safe general information or help, and measure refusals and false refusals across contexts and languages.
Detailed explanation
A
IncorrectReturn part of the dangerous procedure
The refusal boundary is defeated.
B
IncorrectDisclose every hidden safety rule
The details can aid attacks.
C
CorrectState the boundary briefly and provide safe general information or a professional resource
Users receive a safe next step.
D
CorrectEvaluate refusal, false refusal, paraphrase, context, and language cases
Safety and usefulness are measured.
Check official documentation
Compare the following points with the official documentation.
Path & Pass is an independent learning site and is not affiliated with any certification body. Questions are original practice content based on published objectives.