Azure AI Fundamentals (AI-901)

How many principles does Microsoft's Responsible AI Standard define, and what are they?

• ASix — fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability
• BFive — fairness, robustness, privacy, explainability, and human oversight of automated decisions
• CFour — accountability, transparency, safety, and lawfulness, applied across the whole model lifecycle
• DSeven — fairness, safety, privacy, inclusiveness, transparency, accountability, and sustainability

A team must keep prompt processing inside the EU. Which deployment type meets that requirement?

• AData Zone Standard — processing stays within the Microsoft-defined EU data zone
• BGlobal Standard — processing follows the region where the resource was created
• CDeveloper — processing is pinned to the geography of the fine-tuned model's origin
• DGlobal Batch — processing is deferred, so requests are queued inside the caller's region

What do you supply when creating a prompt agent?

• AInstructions, a model selection and tools, with Foundry running the agent for you
• BA container image plus a manifest describing the endpoint the runtime should expose
• CA fine-tuned model and a labelled evaluation set for the runtime to score against
• DAn orchestration graph describing every step the agent may take through its tools

Which three methods can field extraction use to produce a field value?

• AExtract, classify and generate — taking values verbatim, categorising, or creating new text
• BExtract, validate and enrich — taking values verbatim, checking them, then adding context
• CDetect, label and normalise — locating spans, tagging them, then converting to canonical form
• DParse, map and merge — reading the structure, aligning it to a schema, then consolidating