McAfee Secure

Exam Code: DP-800

Exam Name: Developing AI-Enabled Database Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: SQL AI Developer Associate

Microsoft DP-800 Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

101 Questions & Answers with Testing Engine

"Developing AI-Enabled Database Solutions Exam", also known as DP-800 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date DP-800 Exam Engine. Your DP-800 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft DP-800 Practice Test 68

Question 670

Queries filter millions of product rows by a JSON property and repeatedly scan payloads. Which preventive practice BEST addresses the underlying risk over time?

  1. Reject unknown fields and bound payload size before committing.
  2. Benchmark maintenance overhead and verify deployment-specific feature support.
  3. Do not treat allocation order as committed business transaction order.
  4. Keep statistics and partition maintenance procedures current.

Explanation

Correct Answer

B:

Benchmark maintenance overhead and verify deployment-specific feature support. This guards against recurrence because unindexed JSON path predicates can incur expensive row-by-row parsing. The direct remediation is: Evaluate supported JSON index or indexed computed-column path for that SQL platform and version.

Incorrect Answers

A:

Reject unknown fields and bound payload size before committing. The current task instead requires this action: benchmark maintenance overhead and verify deployment-specific feature support.

C:

Do not treat allocation order as committed business transaction order.

D:

Keep statistics and partition maintenance procedures current. The current task instead requires this action: benchmark maintenance overhead and verify deployment-specific feature support.

Question 671

An app embeds user queries with one model but documents with another and similarity search becomes erratic. What should the SQL developer do FIRST to address the specific issue?

  1. Use bounded exponential backoff, batching where supported and idempotent writes.
  2. Update the external model configuration through a validated change procedure.
  3. Use compatible embedding model/version for both corpus and query vectors.
  4. Prefer a model and API mode supporting validated structured output and enforce schema checks.

Explanation

Correct Answer

C:

Use compatible embedding model/version for both corpus and query vectors. This targets the primary issue because vectors from different embedding spaces are not reliably comparable by distance. Verification: A gold query set ranks known relevant chunks consistently after alignment. Prevent recurrence: Audit query and corpus model IDs on every release.

Incorrect Answers

A:

External model quotas and transient errors require controlled retry behavior.

B:

External endpoint configuration can become stale independently of database code.

D:

Unconstrained prose is fragile to parse and can omit required keys.

Question 672

A fraud team seeks a two-hop relationship among accounts, devices and shared addresses. Which test result BEST demonstrates that the intended solution is working?

  1. A sample network returns the known minimal path when explicitly requested.
  2. Injected errors leave balances unchanged and propagate a useful error.
  3. The query returns one expected latest event for each sensor.
  4. Known two-hop links are retrieved without unrelated reverse links.

Explanation

Correct Answer

D:

Known two-hop links are retrieved without unrelated reverse links. This is the decisive observable evidence after implementing use MATCH with explicit edge direction and aliases for the required path. It tests the underlying requirement: Reversing graph edge direction can change which relationships match.

Incorrect Answers

A:

A sample network returns the known minimal path when explicitly requested. However, that result would not establish the required outcome: known two-hop links are retrieved without unrelated reverse links.

B:

Injected errors leave balances unchanged and propagate a useful error. However, that result would not establish the required outcome: known two-hop links are retrieved without unrelated reverse links.

C:

The query returns one expected latest event for each sensor. However, that result would not establish the required outcome: known two-hop links are retrieved without unrelated reverse links.

Question 673

A legal assistant must answer from a restricted set of client-specific contract clauses. Which technical principle BEST explains the relevant design decision?

  1. RAG is appropriate only if the retriever enforces contractual permissions.
  2. Numeric reconciliation is a structured calculation, not a reason to hallucinate an answer from documents.
  3. Language model fluency does not establish numerical correctness.
  4. Transport errors must be separated from valid inference responses.

Explanation

Correct Answer

A:

RAG is appropriate only if the retriever enforces contractual permissions. This explains why the corrective action is to use tenant-trimmed retrieval with citations and a strict abstention policy. Confirm the diagnosis when a cross-client negative test yields no unauthorized quoted text.

Incorrect Answers

B:

Numeric reconciliation is a structured calculation, not a reason to hallucinate an answer from documents. The actual technical constraint is that RAG is appropriate only if the retriever enforces contractual permissions.

C:

Language model fluency does not establish numerical correctness. The actual technical constraint is that RAG is appropriate only if the retriever enforces contractual permissions.

D:

Transport errors must be separated from valid inference responses. The actual technical constraint is that RAG is appropriate only if the retriever enforces contractual permissions.

Question 674

A dashboard needs all line item identifiers aggregated into one JSON array per order. Following remediation, what safeguard BEST prevents a repeat of this specific failure?

  1. Index frequently filtered paths only after measured need.
  2. Specify ordering when supported and required by consumers.
  3. Set sensible MAXRECURSION and detect cycles.
  4. Limit traversal depth and track source/target referential rules.

Explanation

Correct Answer

B:

Specify ordering when supported and required by consumers. This guards against recurrence because JSON_ARRAYAGG can collect multiple row values into a JSON array. The direct remediation is: Use JSON_ARRAYAGG grouped by the order identifier where supported.

Incorrect Answers

A:

Index frequently filtered paths only after measured need. The current task instead requires this action: specify ordering when supported and required by consumers.

C:

Set sensible MAXRECURSION and detect cycles. The current task instead requires this action: specify ordering when supported and required by consumers.

D:

Limit traversal depth and track source/target referential rules. The current task instead requires this action: specify ordering when supported and required by consumers.

Question 675

A stored procedure integrates with a DAB endpoint but unit tests pass while the API returns wrong roles. Which technical action BEST resolves the primary problem?

  1. Add SQL unit tests for accepted and rejected data cases before deployment.
  2. Use protected branch triggers, environment approval gates and deployment-specific identities.
  3. Add integration tests across database permissions, DAB configuration and HTTP behavior.
  4. Use feature branches and resolve merge conflicts with SQL semantic review and tests.

Explanation

Correct Answer

C:

Add integration tests across database permissions, DAB configuration and HTTP behavior. This targets the primary issue because cross-system contracts can fail even when individual SQL procedures are correct. Verification: Authorized and denied endpoint tests match the intended policy. Prevent recurrence: Keep isolated test credentials and avoid production data.

Incorrect Answers

A:

Schema compilation alone does not prove business rules are correct.

B:

Using protected branch triggers, environment approval gates and deployment-specific identities addresses a different database behavior; the relevant distinction is that add integration tests across database permissions, DAB configuration and HTTP behavior.

D:

A text merge can compile but still change business behavior.

Question 676

A developer enables broad verbose SQL tracing and production throughput falls sharply. Which test result BEST demonstrates that the intended solution is working?

  1. A test UPDATE appears in protected audit storage with correct identity.
  2. Cross-tenant nested queries return no unauthorized records in negative tests.
  3. An unapproved identity is denied even from an allowed network.
  4. Selected security events are captured with tolerable overhead.

Explanation

Correct Answer

D:

Selected security events are captured with tolerable overhead. This is the decisive observable evidence after implementing use targeted audit groups and evaluate performance and retention costs. It tests the underlying requirement: High-volume blanket logging can impact throughput and reveal sensitive inputs.

Incorrect Answers

A:

A test UPDATE appears in protected audit storage with correct identity. However, that result would not establish the required outcome: selected security events are captured with tolerable overhead.

B:

Cross-tenant nested queries return no unauthorized records in negative tests. However, that result would not establish the required outcome: selected security events are captured with tolerable overhead.

C:

An unapproved identity is denied even from an allowed network. However, that result would not establish the required outcome: selected security events are captured with tolerable overhead.

Question 677

An Azure-hosted worker stores a SQL username and password in its deployment manifest. Which technical principle BEST explains the relevant design decision?

  1. Managed identities remove a static SQL password from the deployed application configuration.
  2. Azure identity authentication and database authorization are distinct checks.
  3. Least-privilege object permissions reduce exposed DDL and direct data paths.
  4. API generation does not automatically imply that exposed data is appropriately authorized.

Explanation

Correct Answer

A:

Managed identities remove a static SQL password from the deployed application configuration. This explains why the corrective action is to use Microsoft Entra-based managed identity authentication where supported. Confirm the diagnosis when the worker connects with an approved identity and no stored SQL password.

Incorrect Answers

B:

Azure identity authentication and database authorization are distinct checks. The actual technical constraint is that managed identities remove a static SQL password from the deployed application configuration.

C:

Least-privilege object permissions reduce exposed DDL and direct data paths. The actual technical constraint is that managed identities remove a static SQL password from the deployed application configuration.

D:

API generation does not automatically imply that exposed data is appropriately authorized. The actual technical constraint is that managed identities remove a static SQL password from the deployed application configuration.

Question 678

A directory matches transposed contact first names such as 'Jonh' versus 'John'. Following remediation, what safeguard BEST prevents a repeat of this specific failure?

  1. Use supplemental identifiers and human approval for merges.
  2. Do not use name similarity to bypass authentication.
  3. Check engine support and trim blank tokens before inserting.
  4. Specify ordering when supported and required by consumers.

Explanation

Correct Answer

B:

Do not use name similarity to bypass authentication. This guards against recurrence because jaro-Winkler-style measures can treat transpositions and prefixes differently from raw edits. The direct remediation is: Test JARO_WINKLER_DISTANCE or related supported fuzzy metric against a reference dataset.

Incorrect Answers

A:

Use supplemental identifiers and human approval for merges. The current task instead requires this action: do not use name similarity to bypass authentication.

C:

Check engine support and trim blank tokens before inserting. The current task instead requires this action: do not use name similarity to bypass authentication.

D:

Specify ordering when supported and required by consumers. The current task instead requires this action: do not use name similarity to bypass authentication.

Question 679

A system splits legal clauses every fixed 50 characters, separating negation words from conditions. What is the MOST appropriate immediate corrective approach?

  1. Split content by semantically meaningful sections within model token limits.
  2. Exclude confidential credential fields and embed only approved relevant content.
  3. Use structure-aware chunking and sensible overlap around boundaries.
  4. Use compatible embedding model/version for both corpus and query vectors.

Explanation

Correct Answer

C:

Use structure-aware chunking and sensible overlap around boundaries. This targets the primary issue because splitting within essential context can reverse meaning and reduce retrieval accuracy. Verification: A sampled retrieval retains the complete exclusion condition. Prevent recurrence: Review boundary cases during chunker changes.

Incorrect Answers

A:

Split content by semantically meaningful sections within model token limits. The relevant technical condition instead is that use structure-aware chunking and sensible overlap around boundaries.

B:

Embedding inputs can expose or encode sensitive data without adding retrieval value.

D:

Vectors from different embedding spaces are not reliably comparable by distance.

Question 680

A SQL AI integration needs a managed definition for an approved text embedding endpoint. What observed evidence would BEST confirm successful remediation?

  1. An unauthorized principal cannot call the model while approved jobs can.
  2. The tiered approach meets held-out accuracy with lower cost per accepted result.
  3. All new embeddings insert without dimension errors and search quality is rechecked.
  4. An authorized test generates an embedding with the expected size.

Explanation

Correct Answer

D:

An authorized test generates an embedding with the expected size. This is the decisive observable evidence after implementing create an external model object with supported endpoint, model identifiers and secure credential strategy. It tests the underlying requirement: A database model definition binds SQL AI operations to a specific external provider configuration.

Incorrect Answers

A:

An unauthorized principal cannot call the model while approved jobs can. However, that result would not establish the required outcome: an authorized test generates an embedding with the expected size.

B:

The tiered approach meets held-out accuracy with lower cost per accepted result. However, that result would not establish the required outcome: an authorized test generates an embedding with the expected size.

C:

All new embeddings insert without dimension errors and search quality is rechecked. However, that result would not establish the required outcome: an authorized test generates an embedding with the expected size.

Developing AI-Enabled Database Solutions (DP-800) Practice Tests