McAfee-Secured Website

Exam Code: AB-100

Exam Name: Agentic AI Business Solutions Architect

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Agentic AI Business Solutions Architect

100% Microsoft AB-100 Guaranteed Success With Testking Training Materials

Get Real AB-100 Exam Questions & Expert Verified Answers as Experienced on the Real Exam!

129 Questions & Answers with Testing Engine

"Agentic AI Business Solutions Architect Exam", also known as AB-100 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AB-100 Exam Engine. Your AB-100 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 AB-100 Practice Test 55

Question 540

During a staged review, the engineering team demonstrates a solution based on this design: Apply user-prompt attack defenses and keep credentials inaccessible to the agent's tool context. Auditors want proof of hostile direct user instruction, not merely an architecture diagram. Which test gives that proof?

  1. Submit direct jailbreak attempts and ensure no secrets or privileged tools are exposed
  2. Change a connector destination to an unapproved host and verify route and release gates block it
  3. Replay forged tool output that tries to alter an account and require the service to reject it
  4. Inject hostile instructions into retrieved PDFs and verify no unauthorized external call

Explanation

Correct Answer

A:

Direct prompt-attack tests validate resistance. This directly addresses the scenario's priority: hostile direct user instruction.

Incorrect Answers

B:

Negative egress checks validate endpoint integrity.

C:

Tool-response trust testing detects indirect injection.

D:

A blocked injection plus denied egress demonstrates layered protection.

Question 541

An external order API includes unexpected text instructing the model to cancel the customer's account. Which design choice should be adopted for this scenario?

  1. Lock integrations to reviewed endpoints and approve changes to their service destinations
  2. Validate untrusted tool-response schemas and deny any attempt to elevate response text to commands
  3. Treat retrieved documents as untrusted, inspect prompt attacks and restrict outbound tool actions
  4. Apply user-prompt defenses with protected credentials and independent tool authorization

Explanation

Correct Answer

B:

A successful API call does not make all content in its payload authoritative commands. This directly addresses the scenario's priority: poisoned tool output.

Incorrect Answers

A:

Destination integrity is essential for secure agent tool execution.

C:

External documents cannot grant new data-export privileges.

D:

User messages never elevate to trusted system identity merely by claiming authority.

Question 542

The deployment board has implemented this recommendation: Validate tool-response schemas and treat arbitrary returned text as lower-trust data rather than execution instructions. Because poisoned tool output is an explicit business requirement, which measurable acceptance result is the most direct demonstration of success?

  1. Submit direct jailbreak attempts and ensure no secrets or privileged tools are exposed
  2. Replay forged tool output that tries to alter an account and require the service to reject it
  3. Inject hostile instructions into retrieved PDFs and verify no unauthorized external call
  4. Change a connector destination to an unapproved host and verify route and release gates block it

Explanation

Correct Answer

B:

Tool-response trust testing detects indirect injection. This directly addresses the scenario's priority: poisoned tool output.

Incorrect Answers

A:

Direct prompt-attack tests validate resistance. That is an appropriate response to hostile direct user instruction, but it does not directly satisfy the described priority of poisoned tool output.

C:

A blocked injection plus denied egress demonstrates layered protection.

D:

Negative egress checks validate endpoint integrity.

Question 543

A malicious configuration change points an approved agent tool toward a lookalike external endpoint that captures customer records. What is the most defensible technical approach for this business case?

  1. Apply user-prompt defenses with protected credentials and independent tool authorization
  2. Validate untrusted tool-response schemas and deny any attempt to elevate response text to commands
  3. Lock integrations to reviewed endpoints and approve changes to their service destinations
  4. Treat retrieved documents as untrusted, inspect prompt attacks and restrict outbound tool actions

Explanation

Correct Answer

C:

Destination integrity is essential for secure agent tool execution. This directly addresses the scenario's priority: connector endpoint substitution.

Incorrect Answers

A:

User messages never elevate to trusted system identity merely by claiming authority.

B:

A successful API call does not make all content in its payload authoritative commands.

D:

External documents cannot grant new data-export privileges.

Question 544

An implementation review confirms the agent design now includes the following: Pin approved connector destinations and review endpoint/configuration changes before release. A skeptical business owner asks for objective evidence about connector endpoint substitution. Which result should be supplied?

  1. Submit direct jailbreak attempts and ensure no secrets or privileged tools are exposed
  2. Change a connector destination to an unapproved host and verify route and release gates block it
  3. Replay forged tool output that tries to alter an account and require the service to reject it
  4. Inject hostile instructions into retrieved PDFs and verify no unauthorized external call

Explanation

Correct Answer

B:

Negative egress checks validate endpoint integrity. This directly addresses the scenario's priority: connector endpoint substitution.

Incorrect Answers

A:

Direct prompt-attack tests validate resistance. That is an appropriate response to hostile direct user instruction, but it does not directly satisfy the described priority of connector endpoint substitution.

C:

Tool-response trust testing detects indirect injection.

D:

A blocked injection plus denied egress demonstrates layered protection.

Question 545

A deployed custom model's weights may be replaced by an unreviewed artifact in a shared storage container. What is the most defensible technical approach for this business case?

  1. Authenticate approved model artifacts and restrict version promotion permissions
  2. Constrain runtime tool privileges to the minimum capability needed for inference
  3. Validate dependencies and block unresolved vulnerable runtime components before promotion
  4. Review training-data provenance, storage, service roles and transfer controls

Explanation

Correct Answer

A:

Model-integrity controls prevent unauthorized replacement. This directly addresses the scenario's priority: unauthorized model update.

Incorrect Answers

B:

Unused high-impact capabilities expand the attack surface.

C:

Output accuracy does not remove software supply-chain risk.

D:

Model training creates a separate sensitive-data movement path.

Question 546

The deployment board has implemented this recommendation: Require signed or verified model artifacts and promotion permission checks on the model registry. Because unauthorized model update is an explicit business requirement, which measurable acceptance result is the most direct demonstration of success?

  1. Run a candidate image vulnerability scan with controlled remediation or approved exception handling
  2. Probe prohibited tools and confirm the runtime cannot invoke unauthorized actions
  3. Tamper with an artifact and verify signatures or registry permission checks reject deployment
  4. Trace approved training data from ingestion through job logs to output model artifacts

Explanation

Correct Answer

C:

Signature/lineage validation demonstrates integrity. This directly addresses the scenario's priority: unauthorized model update.

Incorrect Answers

A:

Artifact-based remediation validates dependency security.

B:

Capability minimization verifies safe scope. That is an appropriate response to model capability escalation, but it does not directly satisfy the described priority of unauthorized model update.

D:

Full training lifecycle evidence validates protection.

Question 547

Engineers use a hosted tuning service but have not reviewed whether confidential support transcripts may leave the approved environment. Which supported design would address the root of this requirement?

  1. Review training-data provenance, storage, service roles and transfer controls
  2. Constrain runtime tool privileges to the minimum capability needed for inference
  3. Validate dependencies and block unresolved vulnerable runtime components before promotion
  4. Authenticate approved model artifacts and restrict version promotion permissions

Explanation

Correct Answer

A:

Model training creates a separate sensitive-data movement path. This directly addresses the scenario's priority: training-data exfiltration.

Incorrect Answers

B:

Unused high-impact capabilities expand the attack surface.

C:

Output accuracy does not remove software supply-chain risk.

D:

Model-integrity controls prevent unauthorized replacement.

Question 548

An implementation review confirms the agent design now includes the following: Validate tuning data residency, retention and authorized service access before starting customization. A skeptical business owner asks for objective evidence about training-data exfiltration. Which result should be supplied?

  1. Trace approved training data from ingestion through job logs to output model artifacts
  2. Probe prohibited tools and confirm the runtime cannot invoke unauthorized actions
  3. Tamper with an artifact and verify signatures or registry permission checks reject deployment
  4. Run a candidate image vulnerability scan with controlled remediation or approved exception handling

Explanation

Correct Answer

A:

Full training lifecycle evidence validates protection. This directly addresses the scenario's priority: training-data exfiltration.

Incorrect Answers

B:

Capability minimization verifies safe scope. That is an appropriate response to model capability escalation, but it does not directly satisfy the described priority of training-data exfiltration.

C:

Signature/lineage validation demonstrates integrity.

D:

Artifact-based remediation validates dependency security.

Question 549

An assistant only needs document summaries, but its model endpoint has been configured with broad external tool invocation powers. Which technical design is most directly justified by the evidence and requirement?

  1. Constrain runtime tool privileges to the minimum capability needed for inference
  2. Review training-data provenance, storage, service roles and transfer controls
  3. Authenticate approved model artifacts and restrict version promotion permissions
  4. Validate dependencies and block unresolved vulnerable runtime components before promotion

Explanation

Correct Answer

A:

Unused high-impact capabilities expand the attack surface. This directly addresses the scenario's priority: model capability escalation.

Incorrect Answers

B:

Model training creates a separate sensitive-data movement path.

C:

Model-integrity controls prevent unauthorized replacement.

D:

Output accuracy does not remove software supply-chain risk.

Question 550

During a staged review, the engineering team demonstrates a solution based on this design: Restrict allowed tool capabilities and runtime permissions to those required for document analysis. Auditors want proof of model capability escalation, not merely an architecture diagram. Which test gives that proof?

  1. Tamper with an artifact and verify signatures or registry permission checks reject deployment
  2. Probe prohibited tools and confirm the runtime cannot invoke unauthorized actions
  3. Trace approved training data from ingestion through job logs to output model artifacts
  4. Run a candidate image vulnerability scan with controlled remediation or approved exception handling

Explanation

Correct Answer

B:

Capability minimization verifies safe scope. This directly addresses the scenario's priority: model capability escalation.

Incorrect Answers

A:

Signature/lineage validation demonstrates integrity.

C:

Full training lifecycle evidence validates protection.

D:

Artifact-based remediation validates dependency security.

Agentic AI Business Solutions Architect (AB-100) Practice Tests