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 38

Question 370

A developer assumes a SQL firewall rule alone authenticates callers to Azure SQL. Which technical action BEST resolves the primary problem?

  1. Use Microsoft Entra-based managed identity authentication where supported.
  2. Require encrypted transport and a verified principal alongside network access restrictions.
  3. Configure distinct least-privilege grants on model and search resources.
  4. Evaluate SQL column-level encryption with explicit key and certificate controls.

Explanation

Correct Answer

B:

Require encrypted transport and a verified principal alongside network access restrictions. This targets the primary issue because network allowlisting does not establish user identity or object authorization. Verification: An unapproved identity is denied even from an allowed network. Prevent recurrence: Keep firewall, identity and database grants under independent review.

Incorrect Answers

A:

Managed identities remove a static SQL password from the deployed application configuration.

C:

Identity permission at one Azure service does not imply access to adjacent resources.

D:

Server-side encryption can decrypt inside SQL with authorized key access, unlike client-controlled Always Encrypted.

Question 371

A product record contains English and Arabic descriptions for different locales. What observed evidence would BEST confirm successful remediation?

  1. Stock changes do not regenerate descriptive vectors yet filters use current quantities.
  2. The tiered approach meets held-out accuracy with lower cost per accepted result.
  3. A bilingual test set retrieves correct permitted items for each locale.
  4. Clause queries retrieve the specific correct paragraph rather than the entire contract.

Explanation

Correct Answer

C:

A bilingual test set retrieves correct permitted items for each locale. This is the decisive observable evidence after implementing select language-appropriate content and model capabilities while retaining locale as metadata. It tests the underlying requirement: Embedding the wrong language source can degrade multilingual retrieval quality.

Incorrect Answers

A:

Stock changes do not regenerate descriptive vectors yet filters use current quantities. However, that result would not establish the required outcome: a bilingual test set retrieves correct permitted items for each locale.

B:

The tiered approach meets held-out accuracy with lower cost per accepted result. However, that result would not establish the required outcome: a bilingual test set retrieves correct permitted items for each locale.

D:

Clause queries retrieve the specific correct paragraph rather than the entire contract. However, that result would not establish the required outcome: a bilingual test set retrieves correct permitted items for each locale.

Question 372

An engineer uses personal GitHub Copilot credentials for a restricted company repository. What is the MOST relevant technical distinction in this situation?

  1. Copilot in Fabric availability can depend on region, capacity and tenant policies.
  2. AI suggestions are proposals, not change-control authorization.
  3. A malicious or compromised MCP server may exfiltrate data or supply unsafe tool behavior.
  4. Personal and enterprise assistants can have different retention and policy controls.

Explanation

Correct Answer

D:

Personal and enterprise assistants can have different retention and policy controls. This explains why the corrective action is to switch to an organization-approved identity and governed Copilot configuration. Confirm the diagnosis when a test session follows enterprise policy without exposing restricted repo content.

Incorrect Answers

A:

Copilot in Fabric availability can depend on region, capacity and tenant policies. The actual technical constraint is that personal and enterprise assistants can have different retention and policy controls.

B:

AI suggestions are proposals, not change-control authorization. The actual technical constraint is that personal and enterprise assistants can have different retention and policy controls.

C:

A malicious or compromised MCP server may exfiltrate data or supply unsafe tool behavior. The actual technical constraint is that personal and enterprise assistants can have different retention and policy controls.

Question 373

A multitenant SQL application sometimes returns another tenant's invoices when a caller forgets a WHERE clause. Which ongoing control is MOST relevant after the issue is fixed?

  1. Prevent arbitrary tenant context spoofing and test privileged access.
  2. Reset pooled connections and validate context on each request.
  3. Protect client-side keys and plan query limitations and rotation.
  4. Scope identity permissions and review token failures.

Explanation

Correct Answer

A:

Prevent arbitrary tenant context spoofing and test privileged access. This guards against recurrence because RLS enforces row filtering in the database rather than relying on every query to remember a filter. The direct remediation is: Implement an appropriate RLS security predicate tied to trusted tenant context.

Incorrect Answers

B:

Reset pooled connections and validate context on each request. The current task instead requires this action: prevent arbitrary tenant context spoofing and test privileged access.

C:

Protect client-side keys and plan query limitations and rotation. The current task instead requires this action: prevent arbitrary tenant context spoofing and test privileged access.

D:

Scope identity permissions and review token failures. The current task instead requires this action: prevent arbitrary tenant context spoofing and test privileged access.

Question 374

An embedding pipeline includes account passwords and credit card numbers in text sent to an external model. What is the MOST appropriate immediate corrective approach?

  1. Keep rapidly changing structured quantities as filterable columns outside the descriptive embedding text.
  2. Exclude confidential credential fields and embed only approved relevant content.
  3. Create an external model object with supported endpoint, model identifiers and secure credential strategy.
  4. Choose and test a multimodal-capable external model with documented input support.

Explanation

Correct Answer

B:

Exclude confidential credential fields and embed only approved relevant content. This targets the primary issue because embedding inputs can expose or encode sensitive data without adding retrieval value. Verification: A sampled outbound model request contains no restricted secrets. Prevent recurrence: Apply column allowlists and pre-inference DLP checks.

Incorrect Answers

A:

Frequent numeric fields add update cost and can weaken semantic relevance.

C:

A database model definition binds SQL AI operations to a specific external provider configuration.

D:

Text-only models cannot directly interpret image content without a supporting vision pipeline.

Question 375

An engineer drops an external model while stored procedures still depend on it. What observed evidence would BEST confirm successful remediation?

  1. An unauthorized principal cannot call the model while approved jobs can.
  2. A sampled outbound model request contains no restricted secrets.
  3. All dependent procedures run in staging against the new model.
  4. A gold query set ranks known relevant chunks consistently after alignment.

Explanation

Correct Answer

C:

All dependent procedures run in staging against the new model. This is the decisive observable evidence after implementing inventory dependent objects and stage a compatible model transition. It tests the underlying requirement: Database objects referring to external model names may fail after removal.

Incorrect Answers

A:

An unauthorized principal cannot call the model while approved jobs can. However, that result would not establish the required outcome: all dependent procedures run in staging against the new model.

B:

A sampled outbound model request contains no restricted secrets. However, that result would not establish the required outcome: all dependent procedures run in staging against the new model.

D:

A gold query set ranks known relevant chunks consistently after alignment. However, that result would not establish the required outcome: all dependent procedures run in staging against the new model.

Question 376

A REST API must execute an approved report stored procedure but not arbitrary SQL. What is the MOST relevant technical distinction in this situation?

  1. Mapping a view can create a governed projection with narrower permissions than base tables.
  2. DAB generates endpoints from explicit configuration rather than automatically exposing every table.
  3. DAB entity projection and permissions determine the API data surface.
  4. DAB can surface declared stored procedures without granting arbitrary query execution.

Explanation

Correct Answer

D:

DAB can surface declared stored procedures without granting arbitrary query execution. This explains why the corrective action is to expose the stored procedure via a restricted DAB stored-procedure entity with execute rights. Confirm the diagnosis when an authorized request runs the report and rejects unauthorized procedure calls.

Incorrect Answers

A:

Mapping a view can create a governed projection with narrower permissions than base tables. The actual technical constraint is that DAB can surface declared stored procedures without granting arbitrary query execution.

B:

DAB generates endpoints from explicit configuration rather than automatically exposing every table. The actual technical constraint is that DAB can surface declared stored procedures without granting arbitrary query execution.

C:

DAB entity projection and permissions determine the API data surface. The actual technical constraint is that DAB can surface declared stored procedures without granting arbitrary query execution.

Question 377

A DiskANN-style vector index returns low recall after a data distribution change. Which ongoing control is MOST relevant after the issue is fixed?

  1. Monitor quality and refresh indexes after corpus shifts.
  2. Use version-specific supported index documentation.
  3. Tune fusion and metadata filters from labeled evaluations.
  4. Benchmark quality loss and index size before reducing precision.

Explanation

Correct Answer

A:

Monitor quality and refresh indexes after corpus shifts. This guards against recurrence because approximate structures trade memory, speed and recall depending on data and configuration. The direct remediation is: Benchmark supported ANN index tuning and rebuild/retraining options against exact KNN results.

Incorrect Answers

B:

Using version-specific supported index documentation addresses a different database behavior; the relevant distinction is that monitor quality and refresh indexes after corpus shifts.

C:

Tune fusion and metadata filters from labeled evaluations. The current task instead requires this action: monitor quality and refresh indexes after corpus shifts.

D:

Benchmark quality loss and index size before reducing precision. The current task instead requires this action: monitor quality and refresh indexes after corpus shifts.

Question 378

Two concurrent sessions read the same inventory stock and each sell the final item. Which implementation decision BEST addresses this requirement?

  1. Require an expected rowversion predicate and verify rows affected before confirming success.
  2. Use appropriate locking or optimistic concurrency validation to make stock decrement atomic.
  3. Benchmark MAXDOP and cost threshold changes using actual plans and instance guidance.
  4. Compare actual execution plans and Query Store history to identify plan regressions.

Explanation

Correct Answer

B:

Use appropriate locking or optimistic concurrency validation to make stock decrement atomic. This targets the primary issue because a read-then-write race can oversell without a concurrency constraint. Verification: Simultaneous sale tests yield at most the available stock quantity. Prevent recurrence: Use a nonnegative stock CHECK and retry-safe transaction design.

Incorrect Answers

A:

Optimistic concurrency only works if conflicts are detected, not silently overwritten.

C:

Benchmark MAXDOP and cost threshold changes using actual plans and instance guidance. The relevant technical condition instead is that use appropriate locking or optimistic concurrency validation to make stock decrement atomic.

D:

Access path and estimated cardinality changes can explain runtime degradation.

Question 379

A support analyst must read a view but not query its underlying private tables directly. Which measurable outcome BEST verifies that the requirement is satisfied?

  1. The identity creates an approved test object but cannot alter production schemas.
  2. Two test principals receive appropriately different display values.
  3. The test principal reads the approved view and fails direct private-table SELECT.
  4. A privileged SQL-only session cannot read plaintext while an approved client decrypts.

Explanation

Correct Answer

C:

The test principal reads the approved view and fails direct private-table SELECT. This is the decisive observable evidence after implementing grant SELECT on the approved view and validate ownership-chain and execution context behavior. It tests the underlying requirement: Exposing a constrained view can limit data surface without broad base-table rights.

Incorrect Answers

A:

The identity creates an approved test object but cannot alter production schemas. However, that result would not establish the required outcome: the test principal reads the approved view and fails direct private-table SELECT.

B:

Two test principals receive appropriately different display values. However, that result would not establish the required outcome: the test principal reads the approved view and fails direct private-table SELECT.

D:

A privileged SQL-only session cannot read plaintext while an approved client decrypts. However, that result would not establish the required outcome: the test principal reads the approved view and fails direct private-table SELECT.

Question 380

A product-description table updates frequently and embeddings must refresh asynchronously without heavy write transactions. Which technical reason BEST explains the proposed correction?

  1. The binding can detect changes and invoke processing without polling arbitrary full tables.
  2. Splitting within essential context can reverse meaning and reduce retrieval accuracy.
  3. Text-only models cannot directly interpret image content without a supporting vision pipeline.
  4. Asynchronous change-based updates decouple embedding cost from OLTP commit latency.

Explanation

Correct Answer

D:

Asynchronous change-based updates decouple embedding cost from OLTP commit latency. This explains why the corrective action is to use Change Tracking or CDC to identify changed records and process them in a background worker. Confirm the diagnosis when updated products receive new embeddings within the freshness SLA.

Incorrect Answers

A:

The binding can detect changes and invoke processing without polling arbitrary full tables. The actual technical constraint is that asynchronous change-based updates decouple embedding cost from OLTP commit latency.

B:

Splitting within essential context can reverse meaning and reduce retrieval accuracy. The actual technical constraint is that asynchronous change-based updates decouple embedding cost from OLTP commit latency.

C:

Text-only models cannot directly interpret image content without a supporting vision pipeline. The actual technical constraint is that asynchronous change-based updates decouple embedding cost from OLTP commit latency.

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