Run adversarial retrieval tests against the booking tools | AIGP
7-day money-back guarantee — full refund within 7 days of purchase if you've completed under 20% of the questions. See pricing →
Certifications Tools Flashcards Career Paths Exam Guides Blog Pricing For Teams About

Language

✓ EnglishDeutschEspañolFrançaisPortuguês
Check readiness — free →

Run adversarial retrieval tests against the booking tools: Which evidence best resolves the uncertainty?

AIGP Understanding How to Govern AI Deployment and Use Medium

Adversarial retrieval tests show whether untrusted text remains data and consequential tools remain controlled.

The question

An EU travel-support assistant retrieves hotel policies and can issue limited booking changes. Retrieved documents may contain embedded instructions. Under the organization’s internal approval process, applicable EU AI Act obligations must be met. Approval requires evidence that hostile retrieved text remains data, booking tools stay restricted, consequential changes require approval, and actions are logged. Which evidence best resolves the uncertainty?

Preparing for AIGP? Take the free 5-min readiness quiz →

  1. Run adversarial retrieval tests against the booking tools. ✓
    Adversarial tests directly exercise prompt injection, tool restrictions, approval gates, and observable actions under realistic conditions.
  2. Review ordinary answer accuracy.
    Ordinary answer quality does not show whether hostile retrieved text can redirect booking actions.
  3. Ask support staff to rate whether retrieved policies look trustworthy in routine use.
    Staff impressions do not test whether untrusted text changes tool calls, bypasses approval, or produces the required logs.
  4. Inspect supplier architecture and security documentation for the retrieval and booking integration.
    Documentation can describe intended controls but cannot demonstrate behavior when retrieved content contains adversarial instructions.
The trap
Test hostile retrieved content against real tools; documentation or ordinary accuracy cannot establish authorization safety.

How to remember it

Adversarial retrieval tests show whether untrusted text remains data and consequential tools remain controlled.

How many of these would you get right?

One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.

Test your AIGP readiness — free

More Understanding How to Govern AI Deployment and Use questions

Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use · Every answer, right and wrong, comes with its own explanation.