Microsoft AI-103 Practice Test 22
Question 210
A document index uses metadata-based access filters, but a query forgets to send the tenant filter. What enforcement is missing?
- Require mandatory authorization filters at the search service boundary, not as an optional prompt instruction.
- Let each calling application choose whether to include a tenant filter.
- Filter only the final answer after retrieving documents from every tenant.
- Rely on a system prompt saying not to mention other tenants.
Explanation
Correct Answer
A:An omitted filter can expose unauthorized text to the model; server-side enforcement should reject unscoped searches.
Incorrect Answers
B:An optional client filter can be omitted, allowing cross-tenant documents into retrieval results.
C:The unauthorized source has already entered processing before answer-level filtering.
D:A prompt instruction about keeping tenants separate is not an enforceable search authorization filter. The retrieval tool must apply trusted tenant and principal claims to every indexed-document query before returning snippets.
Question 211
An agent forgets an earlier customer approval after using several tools. What context must be retained?
- Minimize conversation memory and redact secrets before persistence or replay
- Maintain a short ordered conversation and approval state tied to the task
- Expire stale task context and retrieve current policy before sensitive decisions
- Discard the recorded customer approval after every tool call and ask the model to infer consent from tone.
Explanation
Correct Answer
B:Relevant conversation state keeps authorization and progress coherent across tool cycles. This directly supports the need to retain task-scoped user decisions during turns, unlike a mechanism primarily intended to keep unnecessary secrets out of agent memory.
Incorrect Answers
A:Retaining only necessary state reduces exposure of sensitive credentials in future prompts. The function serves to keep unnecessary secrets out of agent memory, whereas the actual requirement is to retain task-scoped user decisions during turns.
C:Short-lived task facts should not override newer authoritative policy evidence. Here the needed outcome is to retain task-scoped user decisions during turns; an option designed to limit stale memory from overruling current rules would not directly provide it.
D:Inferred sentiment is not an auditable approval state. Carry the verified approval, action scope and ordered task context through the current session.
Question 212
A deployment workflow includes credentials copied into pipeline YAML. What safer authentication change should be made?
- Use workload identity or managed federation rather than embedded pipeline secrets
- Run post-deployment connectivity and permission smoke tests for critical tools
- Run a representative evaluation suite with defined failure thresholds before promotion
- Publish the same pipeline YAML and infrastructure template to Git without removing embedded client secrets or rotating the exposed credentials.
Explanation
Correct Answer
A:Federated or managed identities avoid long-lived credentials committed to repositories. The decisive difference is whether the design can remove static deployment pipeline secrets; the competing function serves to validate actual deployed tool dependencies.
Incorrect Answers
B:Build tests cannot validate environment-specific tool networking and authorization. Here the needed outcome is to remove static deployment pipeline secrets; an option designed to validate actual deployed tool dependencies would not directly provide it.
C:A measurable evaluation gate prevents known regressions from reaching users. Its primary purpose is to test model behavior before promotion; the stated constraint instead demands a way to remove static deployment pipeline secrets.
D:Version-controlled infrastructure improves repeatability but does not secure secrets already embedded in pipeline files. Replace long-lived plaintext credentials with a supported workload identity or secured secret reference, then rotate exposed values.
Question 213
An app generates alt text for a decorative separator image used only for visual styling. What should its accessibility behavior do?
- Describe meaningful visual content directly without redundant introductory phrases
- Generate a verbose description of the decorative separator for every screen-reader encounter.
- Describe the decorative separator at length in the alt attribute and repeat the same text in an adjacent paragraph.
- Use an empty decorative alt attribute rather than describing irrelevant decoration
Explanation
Correct Answer
D:Decorative images should not add noise for screen-reader navigation. Here the relevant capability must let the team omit noninformative decoration from spoken content. An alternative targeting write concise direct alt descriptions solves another problem.
Incorrect Answers
A:Screen readers already convey the image role, so descriptive content should carry the information. This helps when a team needs to write concise direct alt descriptions; it does not establish how to omit noninformative decoration from spoken content.
B:A purely decorative separator conveys no necessary content, so verbose alternative text creates noise. The appropriate web alternative is an empty alt attribute.
C:Pure decoration need not create additional screen-reader speech. A decorative image can use an empty alt attribute rather than a lengthy redundant description.
Question 214
An input language is unknown and a translation request is frequently routed incorrectly. What preliminary step helps?
- Force every unknown-language request through English speech recognition before translating
- Detect the source language before choosing the translation direction
- Use Azure Translator with a managed terminology strategy where supported
- Request faithful translation only and validate that no extra commentary was added
Explanation
Correct Answer
B:Language identification prevents using the wrong source and target settings. Here the relevant capability must let the team select the correct source language automatically. An alternative targeting protect exact business values during translation solves another problem.
Incorrect Answers
A:An unrecognized language cannot be safely transcribed by assuming English. Detect or establish the source language before selecting the appropriate recognition and translation path.
C:A dedicated translation flow can provide multilingual conversion while preserving approved terminology. That capability is appropriate for translate support text with consistent terminology, not this separate task of finding how to select the correct source language automatically.
D:Translation output should preserve meaning and scope rather than introduce advisory material. Here the needed outcome is to select the correct source language automatically; an option designed to avoid additions when translating legal text would not directly provide it.
Question 215
A vector query performs poorly after document and query embeddings are generated by unrelated models. What consistency rule matters?
- Rebuild only the query vector with a different embedding model while leaving document vectors untouched
- Align embedding output dimension with the index vector field definition
- Use compatible embedding representations for indexed documents and search queries
- Use hybrid keyword and vector retrieval in one query workflow
Explanation
Correct Answer
C:Vector distances are meaningful only when query and document vectors inhabit compatible spaces. This resolves the specific need to keep document and query embeddings compatible, while a measure that helps avoid redundant evidence in hybrid retrieval addresses a different requirement.
Incorrect Answers
A:Vectors from incompatible model spaces cannot be compared as though they share the same geometry. Document and query embeddings need a compatible deployment and version.
B:Vector search requires vectors whose shape agrees with the configured schema. That is relevant to prevent vector field shape mismatches, but the present scenario requires an ability to keep document and query embeddings compatible.
D:Hybrid retrieval covers rare identifiers and semantic descriptions that neither method handles perfectly alone. The function serves to match exact identifiers plus related concepts, whereas the actual requirement is to keep document and query embeddings compatible.
Question 216
A model remembers a customer PIN because full prior chat is always replayed. What safer state design is appropriate?
- Tag each conversation and tool response with stable isolated session and task identifiers
- Minimize conversation memory and redact secrets before persistence or replay
- Store all customer PINs in the agent's durable role description so they remain available across conversations.
- Maintain a short ordered conversation and approval state tied to the task
Explanation
Correct Answer
B:Retaining only necessary state reduces exposure of sensitive credentials in future prompts. The decisive difference is whether the design can keep unnecessary secrets out of agent memory; the competing function serves to isolate concurrent conversation state.
Incorrect Answers
A:Correlation IDs prevent results and memory from leaking across concurrent requests. That capability is appropriate for isolate concurrent conversation state, not this separate task of finding how to keep unnecessary secrets out of agent memory.
C:Putting authentication secrets in long-term agent instructions increases leakage and retention risk. Minimize replayed state and redact sensitive fields.
D:Relevant conversation state keeps authorization and progress coherent across tool cycles. That is relevant to retain task-scoped user decisions during turns, but the present scenario requires an ability to keep unnecessary secrets out of agent memory.
Question 217
Several models share a subscription quota, and a new deployment fails with a throughput allocation error. Which quota dimension should be checked first?
- Inspect available capacity for that model family, deployment type, and region before resizing requests.
- Assume a quota value in an unrelated region is available to this deployment.
- Increase web app instances without checking model throughput entitlements.
- Rotate deployment credentials to clear a throughput allocation error.
Explanation
Correct Answer
A:Model quotas can be scoped by deployment type and location. An unused quota elsewhere does not necessarily authorize the target allocation.
Incorrect Answers
B:Model family and deployment-type quotas can be region-specific and cannot be freely transferred.
C:Adding App Service web workers changes application hosting capacity, not the model deployment's assigned token throughput. Inspect the Foundry model, region, deployment type and TPM allocations before increasing serving capacity.
D:Fresh credentials do not resolve exhausted model family or regional quota.
Question 218
A diagram is essential to instructions, but a one-sentence alt text cannot convey its labeled sequence. What accessible support is appropriate?
- Use only the diagram file name as alternative text and omit its labeled sequence of steps.
- Provide a concise alt summary and a separate detailed equivalent description
- Write context-specific descriptions of each step rather than repeating one generic caption
- Use an empty decorative alt attribute rather than describing irrelevant decoration
Explanation
Correct Answer
B:Complex instructional graphics often need a longer structured text alternative. This directly supports the need to describe complex diagrams beyond short alt text, unlike a mechanism primarily intended to avoid speculative attributes in alt text.
Incorrect Answers
A:A file name does not communicate the procedure shown by the diagram. Keep concise alt text and provide a fuller equivalent description of the essential steps.
C:Per-image alternatives let nonvisual users follow the sequence. Here the needed outcome is to describe complex diagrams beyond short alt text; an option designed to distinguish meaningful steps across several images would not directly provide it.
D:Decorative images should not add noise for screen-reader navigation. Its primary purpose is to omit noninformative decoration from spoken content; the stated constraint instead demands a way to describe complex diagrams beyond short alt text.
Question 219
A multilingual chatbot changes currency amounts while translating invoices. What validation is required?
- Convert currency quantities to convenient rounded values during translation
- Detect the source language before choosing the translation direction
- Preserve and compare numeric, currency and identifier tokens across translations
- Use Azure Translator with a managed terminology strategy where supported
Explanation
Correct Answer
C:Translation must not silently alter financial facts or reference numbers. This directly supports the need to protect exact business values during translation, unlike a mechanism primarily intended to avoid additions when translating legal text.
Incorrect Answers
A:A translator must preserve amounts and units unless the workflow separately authorizes conversion. Silent rounding can change financial obligations in the invoice.
B:Language identification prevents using the wrong source and target settings. That capability is appropriate for select the correct source language automatically, not this separate task of finding how to protect exact business values during translation.
D:A dedicated translation flow can provide multilingual conversion while preserving approved terminology. Its primary purpose is to translate support text with consistent terminology; the stated constraint instead demands a way to protect exact business values during translation.
Question 220
A hybrid result set is dominated by near-duplicate snippets from one manual. What post-retrieval processing helps?
- Raise top-k while allowing the same manual to occupy every result slot
- Use compatible embedding representations for indexed documents and search queries
- Deduplicate or diversify ranked results while retaining strongest evidence
- Align embedding output dimension with the index vector field definition
Explanation
Correct Answer
C:Source diversity and duplicate filtering improve usable coverage for grounded generation. The practical requirement is to avoid redundant evidence in hybrid retrieval; a solution built to match exact identifiers plus related concepts would not meet that same need.
Incorrect Answers
A:Increasing result count does not address redundancy from one document. Diversification or grouping by source is needed to improve coverage of different evidence.
B:Vector distances are meaningful only when query and document vectors inhabit compatible spaces. This helps when a team needs to keep document and query embeddings compatible; it does not establish how to avoid redundant evidence in hybrid retrieval.
D:Vector search requires vectors whose shape agrees with the configured schema. That capability is appropriate for prevent vector field shape mismatches, not this separate task of finding how to avoid redundant evidence in hybrid retrieval.