Free practice tests for certification exams. Know when you're ready.
Original practice questions written from each exam's official documentation. Every answer comes with an explanation, and your progress shows which areas still need work. No sign-up, no paywall.
- 360+ practice questions
- 2 exams live, more on the way
- New questions every week
Microsoft Azure AI
Microsoft Azure
Try one now · AI-103 · Plan and manage AI solutions
A company in Germany is deploying a chat model in Microsoft Foundry. Compliance requires that prompts and responses are processed only inside the EU. Traffic is unpredictable, so the team wants pay-per-token billing and more default quota than a deployment limited to a single geography. Which deployment type should they choose?
Data Zone Standard keeps processing inside a Microsoft-defined data zone such as the EU, bills per token, and gets higher default quota than geography-based types like Standard. It fits EU-only processing with unpredictable traffic.
- A: Global Standard can route traffic to datacenters in any Azure region, so EU-only processing isn't guaranteed.
- C: Regional Provisioned reserves capacity (PTUs) that you pay for whether you use it or not. It suits steady, predictable traffic, not unpredictable pay-per-token use.
- D: Global Batch is for asynchronous jobs with a 24-hour target turnaround, and it can process data in any region.
How it works
-
1 Practice by skill area
Questions follow each exam's official skills outline and weighting, so you spend time where the exam does.
-
2 Learn from every answer
Each question explains why the right answer is right, why the others are wrong, and links to the official documentation behind it.
-
3 Check you're ready
Timed exams and a readiness view show your weak areas before you book the real exam.
Every question is original. These are not real exam questions. How questions are made
What you see after every answer
A real AI-103 question, with what each part teaches you.
AI-103 · Plan and manage AI solutions
A model deployment summarizes incoming customer emails for an operations team. A tester sends an email that contains hidden text telling the model to post a false security alert on the company blog. The team's guardrail uses the default settings. Which change targets this risk?
- A Enable Prompt Shields document attack detection in the guardrail ✓
- B Lower the violence category's blocking threshold to low
- C Turn on protected material detection for text
- D Add groundedness detection to the guardrail
Why this answer is right
Instructions hidden in emails, documents or web pages are document (indirect) prompt attacks. Prompt Shields can detect them, but document attack detection is off by default, so the team has to turn it on in the guardrail.
Why each other option is wrong
- B: Harm-category thresholds look for harmful content such as violence. An embedded instruction to change the model's behavior isn't caught by them.
- C: Protected material detection looks for copyrighted text. It doesn't detect injected instructions.
- D: Groundedness detection checks whether a response is supported by source material. It doesn't detect an attempt to hijack the model.
The official source
- It teaches, not just grades. The explanation says why the answer fits the scenario.
- Your mistakes get explained too. Every wrong option says exactly why it doesn't fit.
- You can check it. Each question links to the Microsoft Learn page it came from.