2026 Actual Microsoft AI-103 Dumps, Updated Exam Questions for Guaranteed Pass

AI-103 free sample questions and answers

Microsoft AI-103 practice quiz with correct answers for search indexing.

1. Question 1

   You are deploying a support agent that enables users to upload photos.

   You need to automatically classify uploaded images for harmful content. The solution must block content based on severity levels.

   What should you do?

   AImplement image moderation.
   

   BEnable prompt shields.
   

   CApply keyword scanning to optical character recognition (OCR) output by using Azure Vision in Foundry Tools.
   

   DUse blocklists.
   

   Correct answer: A

   The correct answer is A. Implement image moderation. Azure AI Content Safety provides image analysis that classifies uploaded images for harmful content, including harm categories such as hate, sexual content, violence, and self-harm. Microsoft's Content Safety overview states that the Analyze Image API scans images for harmful content with multi-severity levels, which directly matches the requirement to automatically classify uploaded photos and block content based on configured severity thresholds.

   Prompt Shields are intended to detect prompt injection and jailbreak-style attacks against generative models, not to classify image harm categories. Keyword scanning OCR output would only detect visible text extracted from the image and would miss visual harm in the image itself. Blocklists can help match known words or custom patterns, but they are not a complete image safety classifier and do not provide the built-in severity-based image harm classification required here. Image moderation is therefore the correct control for user-uploaded photos. Reference topics: Azure AI Content Safety, image moderation, harm categories, severity levels, Foundry guardrails, and responsible AI controls.

   
   

2. Question 2

   You are implementing an agent-based solution in Azure AI Foundry that automates compliance document review. The agent must extract specific entities and structured data from documents, apply company policies to flag risky clauses, and maintain an audit trail for regulatory purposes. You are deciding how to implement entity extraction, policy enforcement, and audit requirements.

   What is the best approach to meet these requirements?

   AUse generative prompting with an LLM configured to output structured JSON, implement policy rules through tool schemas and function-calling constraints, and enable trace logging with provenance metadata for audit trails
   

   BHardcode all entity extraction patterns as regex rules, manually invoke each policy check as separate agent steps, and log all decisions to a simple text file without metadata
   

   CUse a small language model for entity extraction because it is faster, skip structured output formatting to reduce complexity, and implement policy checks through random sampling of documents
   

   DRely only on LLM in-context learning for compliance rules without tool integration, use unstructured log output, and assume audit requirements can be met through email notifications
   

   Correct answer: A

   Correct Answer: The first option is correct. It leverages LLM structured output for reliable entity extraction, uses tool schemas and function-calling constraints to enforce policies programmatically, and implements proper auditing through trace logging and provenance metadata—all aligned with the exam objectives for responsible AI implementation and agent design.

   Why other options are incorrect: Option 2 is brittle (regex patterns fail on document variance), manual steps do not scale, and text-file logging is insufficient for regulatory audits. Option 3 downgrades to a small language model that lacks the reasoning capability for nuanced compliance analysis, skips structured output (defeating the extraction goal), and introduces random sampling which violates compliance rigor. Option 4 ignores tool integration best practices, lacks proper audit instrumentation, and email notifications are not a compliant audit mechanism.

3. Question 3

   You are building a speech processing solution in Microsoft Foundry for a customer support platform.

   The platform will transcribe live phone calls, so that supervisors at your company can view call transcripts and detect issues while the calls are in progress. The call audio will arrive as a continuous stream from the telephony system.

   You need to ensure that the call transcripts appear within only a few seconds of the audio stream.

   What should you do?

   ARun a batch transcription job on recorded audio files.
   

   BUse real-time speech to text to process streaming audio input.
   

   CUse speech translation to generate the transcripts into multiple languages.
   

   DUse text to speech by using a custom neural voice.
   

   Correct answer: B

   The correct answer is B. Use real-time speech to text to process streaming audio input. The scenario requires live transcription from a continuous telephony stream, with transcript text appearing within a few seconds while the call is still in progress. Azure Speech in Foundry Tools real-time speech recognition is specifically intended for immediate transcription scenarios such as call center assistance, dictation, and live meeting captioning. Microsoft's Speech guidance describes real-time speech to text as processing audio input and returning transcriptions in real time, which matches the supervisor monitoring requirement.

   Batch transcription is inappropriate because it processes stored audio files after recording, not an active live stream. Speech translation is used when the primary goal is translating speech into another language, not simply producing live same-language call transcripts. Text to speech performs the reverse operation by generating spoken audio from text and does not transcribe inbound calls. Real-time speech to text provides the low-latency streaming recognition path required for live operational monitoring. Reference topics: Azure Speech in Foundry Tools, real-time speech recognition, streaming audio input, call center transcription, and live captions.

   
   

4. Question 4

   You are creating an image-editing workflow in a Microsoft Foundry project.

   The workflow must meet the following requirements:

   * Ensure that background objects can be removed by applying a mask-based inpainting edit.

   * Preserve the original lighting and style of the edited images.

   * Use the built-in image editing controls, NOT a custom model.

   You need to ensure that image edits apply exclusively inside the masked area.

   How should you configure the workflow?

   AEnable text_to_image mode and a prompt describing the desired background removal.
   

   BSet generation mode to image_variation and provide the original image as a reference.
   

   CEnable image_to_image mode and a high-strength value to regenerate the full image based on the prompt.
   

   DEnable mask_inpainting and supply both the input image and a mask indicating which part of the image to modify.
   

   Correct answer: D

   The correct configuration is D. Enable mask_inpainting and supply both the input image and a mask indicating which part of the image to modify. The requirement is not to generate a new image, but to edit a specific region of an existing image while preserving the surrounding lighting, composition, and style. Azure OpenAI image editing in Microsoft Foundry supports modifying existing images by submitting an input image plus a prompt. For masked edits, the mask explicitly defines the part of the image the model is allowed to change; Microsoft states that the mask parameter defines the area to edit and must match the input image dimensions.

   text_to_image would create a new image from a prompt and cannot guarantee preservation of the original image. image_variation generates related variants rather than targeted removals. image_to_image with high strength can regenerate broader areas and may alter unrelated visual details. Mask-based inpainting is the built-in editing control that limits modification to the selected region. Reference topics: Azure OpenAI image editing, mask inpainting, image edit API, input image, mask parameter, and computer vision image generation workflows.

   
   

5. Question 5

   Your team is deploying a generative AI application to production using Azure AI Foundry. The application ingests large volumes of customer documents daily, performs semantic search for retrieval-augmented generation, and serves responses via REST API endpoints. You need to establish monitoring and cost management controls to prevent budget overruns while ensuring solution reliability.

   Which monitoring and management strategies should you implement? (Select two.)

   AMonitor token consumption metrics and implement rate limiting at the API layer; configure quotas and track cost footprints by model deployment
   

   BEnable trace logging for all model inference requests and set up alerts on safety events and grounding quality metrics
   

   CDisable safety filters to reduce token overhead and increase throughput, then monitor safety incident counts manually
   

   DAvoid monitoring search index health and relevance performance to reduce operational overhead
   

   EUse only client-side rate limiting without server-side quota enforcement, as quotas are only effective for batch processing
   

   Correct answer: A, B

   Correct Answers: The first and second options are both correct. Option 1 directly addresses cost management by tracking token usage and implementing quotas, which is essential for preventing runaway costs. Option 2 implements proper observability through trace logging and monitors both safety and grounding quality, which are core responsible AI and system reliability practices documented in the exam objectives.

   Why other options are incorrect: Option 3 contradicts responsible AI requirements by disabling safety filters. Option 4 ignores a key monitoring requirement—search index health is critical for RAG quality and performance. Option 5 is incorrect because server-side quotas are essential for enforcing rate limits reliably; client-side controls alone are insufficient and applicable to all workloads, not just batch processing.