A. Incorrect.
Azure AI Translator translates text between languages and is not a full ML lifecycle workspace.

B. Correct.
Azure Machine Learning supports building, training, evaluating, registering, and deploying custom ML models.

C. Incorrect.
Azure AI Speech provides speech capabilities, not a general ML training workspace.

D. Incorrect.
Azure Bot Service hosts and connects conversational bots; it does not provide full model training lifecycle management.

A. Incorrect.
Speech recognition converts spoken audio to text and is not used for image labels.

B. Incorrect.
Entity recognition identifies entities in text, not objects or categories in images.

C. Correct.
Image classification assigns one or more labels to an image.

D. Incorrect.
Regression predicts numeric values and is not the right workload for image labels.

A. Incorrect.
Image classification labels the whole image and does not return bounding boxes for each item.

B. Incorrect.
Text-to-speech generates audio from text and is not a vision workload.

C. Incorrect.
Language detection identifies the language of text and does not analyze objects in images.

D. Correct.
Object detection identifies objects and returns locations, commonly as bounding boxes.

A. Correct.
Azure AI Translator translates text between supported languages.

B. Incorrect.
Azure AI Vision analyzes images and visual content.

C. Incorrect.
Document Intelligence extracts structured fields from documents, not language translation.

D. Incorrect.
Azure AI Speech can support speech translation scenarios, but text document translation is handled by Azure AI Translator.

A. Incorrect.
Sentiment analysis evaluates opinion or emotional tone.

B. Incorrect.
Image tagging labels visual content, not text entities.

C. Correct.
Named entity recognition identifies entities such as people, locations, organizations, dates, and quantities in text.

D. Incorrect.
Object detection locates objects in images and does not extract entities from text.

A. Incorrect.
Data ingestion is a data pipeline activity and not a responsible AI principle.

B. Incorrect.
Model serialization saves a model artifact and does not explain AI behavior to users.

C. Incorrect.
Autoscaling changes resource capacity and is not a responsible AI principle.

D. Correct.
Transparency is about making AI systems understandable, clear, and appropriately disclosed.

A. Incorrect.
Higher randomness and no grounding can increase unsupported or inconsistent answers.

B. Correct.
Grounding connects the prompt to trusted retrieved content so the model can answer from approved sources.

C. Incorrect.
A public IP address is a networking resource and does not improve factual grounding.

D. Incorrect.
A virtual network boundary is unrelated to making generated answers use policy documents.

A. Correct.
Azure AI Content Safety can detect harmful user-generated and AI-generated content categories.

B. Incorrect.
Azure DNS hosts DNS zones and records and does not moderate AI content.

C. Incorrect.
Azure Files provides managed file shares, not safety classification for AI text or images.

D. Incorrect.
Azure Load Testing evaluates application performance under load, not content safety.

A. Incorrect.
Speech-to-text transcribes spoken audio into text, which is the opposite direction.

B. Incorrect.
Key phrase extraction identifies important phrases in text and does not generate audio.

C. Correct.
Text-to-speech converts written text into spoken audio.

D. Incorrect.
Image classification labels images and is unrelated to audio generation.

A. Incorrect.
Azure Network Watcher helps monitor and diagnose Azure networking.

B. Incorrect.
Azure Virtual Network provides private networking for Azure resources and does not create conversational assistants.

C. Incorrect.
Azure Monitor Metrics stores numerical telemetry and does not create conversational assistants.

D. Correct.
Microsoft Copilot Studio is used to create and customize copilots with low-code tools and generative AI capabilities.