Palo Alto Networks NetSec-Architect Practice Test 12
Question 110
For an original educational Network Security Architect scenario, consider: A company plans to roll out a new agentic help desk that can initiate account actions, and security engineers need a controlled way to test jailbreak and unsafe instruction handling. Select TWO complementary steps: the targeted correction and its strongest prechange validation.
- Change: Include controlled AI red-team testing with representative misuse cases, prompt injection challenges, safety benchmarks and documented remediation before enabling privileged integrations.
- Treat passing a network port scan as sufficient proof that a chatbot cannot be manipulated by adversarial user instructions.
- Change: Apply enterprise DLP and AI-aware inspection for sensitive prompt and response content, aligning enforcement with sanctioned applications and documented exceptions.
- Check: Trace each model client, inference endpoint, control plane, identity boundary, payload path and failure behavior before selecting the integration method.
- Check: Examine the evaluation plan, attack prompts, guardrail failures, permission scope and retest results for the exact deployed agent release.
Explanation
Correct Answer
A:Adversarial evaluation probes model and application behavior under malicious prompts, while perimeter firewall tests alone do not expose unsafe agent reasoning or tool permissions. Evidence: Examine the evaluation plan, attack prompts, guardrail failures, permission scope and retest results for the exact deployed agent release.
E:Adversarial evaluation probes model and application behavior under malicious prompts, while perimeter firewall tests alone do not expose unsafe agent reasoning or tool permissions. Required intervention: Include controlled AI red-team testing with representative misuse cases, prompt injection challenges, safety benchmarks and documented remediation before enabling privileged integrations.
Incorrect Answers
B:Adversarial evaluation probes model and application behavior under malicious prompts, while perimeter firewall tests alone do not expose unsafe agent reasoning or tool permissions.
C:Change: Apply enterprise DLP and AI-aware inspection for sensitive prompt and response content, aligning enforcement with sanctioned applications and documented exceptions. A company plans to roll out a new agentic help desk that can initiate account actions, and security engineers need a controlled way to test jailbreak and unsafe instruction handling.
D:Check: Trace each model client, inference endpoint, control plane, identity boundary, payload path and failure behavior before selecting the integration method. Adversarial evaluation probes model and application behavior under malicious prompts, while perimeter firewall tests alone do not expose unsafe agent reasoning or tool permissions.
Question 111
A network security architect is reviewing this environment: An employee-facing assistant may leak confidential customer information in generated answers even though users access it through an approved HTTPS endpoint. A proposed shortcut is: Approve every AI message simply because the endpoint uses TLS and the model previously passed a static scan. What should the security architect recommend as the narrowly scoped correction? Select ONE best answer.
- Approve every AI message simply because the endpoint uses TLS and the model previously passed a static scan.
- Select a Prisma AIRS deployment form factor and integration point consistent with who controls the model endpoint, network path and application API.
- Map AI model and application data flows to applicable regulatory obligations and risk controls, then choose architecture and enforcement based on documented requirements.
- Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
Explanation
Correct Answer
D:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Incorrect Answers
A:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
B:Select a Prisma AIRS deployment form factor and integration point consistent with who controls the model endpoint, network path and application API. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:Map AI model and application data flows to applicable regulatory obligations and risk controls, then choose architecture and enforcement based on documented requirements. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 112
A Palo Alto Networks architecture review finds: An employee-facing assistant may leak confidential customer information in generated answers even though users access it through an approved HTTPS endpoint. A design team proposes: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released. Which architecture principle most strongly supports the recommended response? Select ONE best answer.
- AI application security decisions depend on application identity, tenant and data handling policy, not on whether a URL contains the word artificial intelligence.
- Compliance frameworks set obligations for data use and accountability; tool selection alone does not establish that an AI deployment complies with every jurisdiction.
- AI inspection architectures vary between direct application/API integration and network or infrastructure enforcement; a valid design must observe the request/response path it claims to protect.
- AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Explanation
Correct Answer
D:Demonstrable success requires: Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally. This principle explains the proposed architecture, which is: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
Incorrect Answers
A:AI application security decisions depend on application identity, tenant and data handling policy, not on whether a URL contains the word artificial intelligence. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
B:Compliance frameworks set obligations for data use and accountability; tool selection alone does not establish that an AI deployment complies with every jurisdiction. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:AI inspection architectures vary between direct application/API integration and network or infrastructure enforcement; a valid design must observe the request/response path it claims to protect. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 113
The architecture board has not yet authorized configuration changes. Current evidence indicates: An employee-facing assistant may leak confidential customer information in generated answers even though users access it through an approved HTTPS endpoint. Which evidence best tests the architecture diagnosis before approving a disruptive change? Select ONE best answer.
- Examine the evaluation plan, attack prompts, guardrail failures, permission scope and retest results for the exact deployed agent release.
- Review discovered AI applications, associated users, sanctioned status, DLP event samples and enforcement rules for enterprise versus personal accounts.
- Exercise representative inspection failures and regional failover while checking request routes, policy decisions, latency and alerting.
- Verify the actual application integration path, inspection coverage, data classification tests, blocking outcome and telemetry on permitted and rejected prompts.
Explanation
Correct Answer
D:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. If confirmed, the proportionate intervention is: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
Incorrect Answers
A:Examine the evaluation plan, attack prompts, guardrail failures, permission scope and retest results for the exact deployed agent release. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
B:Review discovered AI applications, associated users, sanctioned status, DLP event samples and enforcement rules for enterprise versus personal accounts. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:Exercise representative inspection failures and regional failover while checking request routes, policy decisions, latency and alerting. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 114
The architecture team has implemented an approved correction and must now perform a production-safe acceptance test. Original problem: An employee-facing assistant may leak confidential customer information in generated answers even though users access it through an approved HTTPS endpoint. The operations owner requires positive and negative test evidence and confirmation that legitimate business workflows still function. Which acceptance result demonstrates improved security without breaking authorized workflows? Select ONE best answer.
- Required inference and retrieval paths work under rescheduling, while unrelated workloads cannot access model-management or confidential dataset services.
- The chosen form factor inspects intended AI traffic without creating undisclosed bypass paths or unacceptable application availability risks.
- Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally.
- Approved enterprise AI workflows remain usable and unsanctioned data transfers are logged or prevented per policy.
Explanation
Correct Answer
C:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. The approved design step was: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
Incorrect Answers
A:Required inference and retrieval paths work under rescheduling, while unrelated workloads cannot access model-management or confidential dataset services. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
B:The chosen form factor inspects intended AI traffic without creating undisclosed bypass paths or unacceptable application availability risks. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
D:Approved enterprise AI workflows remain usable and unsanctioned data transfers are logged or prevented per policy. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 115
A design sign-off request identifies the following problem: An employee-facing assistant may leak confidential customer information in generated answers even though users access it through an approved HTTPS endpoint. Which paired security change and validation should close the design review? Select ONE best answer.
- Change: Combine appropriate application identification and cloud-delivered classification features with granular AI application access policy and validation. Verify: Compare actual application logs, cloud classification, affected tenant URLs, allow and block rules and user experience under a test policy.
- Change: Discover and classify AI application use, then apply role-aware sanctioned-app controls and documented data sharing restrictions. Verify: Review discovered AI applications, associated users, sanctioned status, DLP event samples and enforcement rules for enterprise versus personal accounts.
- Change: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released. Verify: Verify the actual application integration path, inspection coverage, data classification tests, blocking outcome and telemetry on permitted and rejected prompts.
- Change: Design an agent boundary with scoped tool permissions, verified user context, untrusted-content isolation and runtime controls for dangerous tool calls. Verify: Test untrusted retrieval content, authorization context, tool invocation logs, deny decisions and the business approval flow for representative high-impact actions.
Explanation
Correct Answer
C:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. The change is successful only when: Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally.
Incorrect Answers
A:Change: Combine appropriate application identification and cloud-delivered classification features with granular AI application access policy and validation.
B:Change: Discover and classify AI application use, then apply role-aware sanctioned-app controls and documented data sharing restrictions. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
D:This option addresses a different security architecture topic (AI agent privilege and tool boundary) rather than the documented AI runtime inspection for prompt and response decision.
Question 116
A network security architect is reviewing this environment: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight. A proposed shortcut is: Approve every AI message simply because the endpoint uses TLS and the model previously passed a static scan. What should the security architect recommend as the narrowly scoped correction? Select ONE best answer.
- Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
- Select a Prisma AIRS deployment form factor and integration point consistent with who controls the model endpoint, network path and application API.
- Approve every AI message simply because the endpoint uses TLS and the model previously passed a static scan.
- Include controlled AI red-team testing with representative misuse cases, prompt injection challenges, safety benchmarks and documented remediation before enabling privileged integrations.
Explanation
Correct Answer
A:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Incorrect Answers
B:Select a Prisma AIRS deployment form factor and integration point consistent with who controls the model endpoint, network path and application API. A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight.
C:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. Use this instead: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
D:Include controlled AI red-team testing with representative misuse cases, prompt injection challenges, safety benchmarks and documented remediation before enabling privileged integrations. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 117
A Palo Alto Networks architecture review finds: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight. A design team proposes: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released. Which architecture principle most strongly supports the recommended response? Select ONE best answer.
- AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
- AI application security decisions depend on application identity, tenant and data handling policy, not on whether a URL contains the word artificial intelligence.
- Compliance frameworks set obligations for data use and accountability; tool selection alone does not establish that an AI deployment complies with every jurisdiction.
- Model scanning assesses risks associated with a model artifact and its behavior before deployment; runtime inspection addresses a different exposure once the model is serving requests.
Explanation
Correct Answer
A:Demonstrable success requires: Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally.
Incorrect Answers
B:AI application security decisions depend on application identity, tenant and data handling policy, not on whether a URL contains the word artificial intelligence. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:Compliance frameworks set obligations for data use and accountability; tool selection alone does not establish that an AI deployment complies with every jurisdiction. A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight.
D:Model scanning assesses risks associated with a model artifact and its behavior before deployment; runtime inspection addresses a different exposure once the model is serving requests. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 118
The architecture board has not yet authorized configuration changes. Current evidence indicates: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight. Which evidence best tests the architecture diagnosis before approving a disruptive change? Select ONE best answer.
- Review discovered AI applications, associated users, sanctioned status, DLP event samples and enforcement rules for enterprise versus personal accounts.
- Trace each model client, inference endpoint, control plane, identity boundary, payload path and failure behavior before selecting the integration method.
- Test synthetic sensitive records, matching classifiers, channel visibility, blocked requests, user notification and exception logging.
- Verify the actual application integration path, inspection coverage, data classification tests, blocking outcome and telemetry on permitted and rejected prompts.
Explanation
Correct Answer
D:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Incorrect Answers
A:Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
B:Trace each model client, inference endpoint, control plane, identity boundary, payload path and failure behavior before selecting the integration method. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:Test synthetic sensitive records, matching classifiers, channel visibility, blocked requests, user notification and exception logging. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 119
The architecture team has implemented an approved correction and must now perform a production-safe acceptance test. Original problem: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight. The operations owner requires positive and negative test evidence and confirmation that legitimate business workflows still function. Which acceptance result demonstrates improved security without breaking authorized workflows? Select ONE best answer.
- Required inference and retrieval paths work under rescheduling, while unrelated workloads cannot access model-management or confidential dataset services.
- Every production AI release has corresponding documented assessment and controls; changed behavior is not silently approved by an older scan.
- The design provides evidence of appropriate classification, access, retention and review for applicable rules without assuming a single universal control.
- Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally.
Explanation
Correct Answer
D:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Incorrect Answers
A:Required inference and retrieval paths work under rescheduling, while unrelated workloads cannot access model-management or confidential dataset services. A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight.
B:Every production AI release has corresponding documented assessment and controls; changed behavior is not silently approved by an older scan. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
C:The design provides evidence of appropriate classification, access, retention and review for applicable rules without assuming a single universal control. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
Question 120
For an original educational Network Security Architect scenario, consider: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight. Select TWO complementary steps: the targeted correction and its strongest prechange validation.
- Change: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
- Check: Review discovered AI applications, associated users, sanctioned status, DLP event samples and enforcement rules for enterprise versus personal accounts.
- Approve every AI message simply because the endpoint uses TLS and the model previously passed a static scan.
- Change: Add a model-assessment stage using appropriate Prisma AIRS AI model scanning and approved remediation or acceptance criteria before production promotion.
- Check: Verify the actual application integration path, inspection coverage, data classification tests, blocking outcome and telemetry on permitted and rejected prompts.
Explanation
Correct Answer
A:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. Acceptance: Sensitive content and malicious AI requests trigger documented controls while approved benign model interactions continue normally.
E:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation. Required intervention: Architect AI runtime protection at a supported enforcement point that can inspect prompts and responses for policy-defined risks before sensitive content is released.
Incorrect Answers
B:Applied case: A public generative AI application sends prompts containing trade secrets to an external model service and needs policy enforcement while requests are in flight.
C:AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.
D:Change: Add a model-assessment stage using appropriate Prisma AIRS AI model scanning and approved remediation or acceptance criteria before production promotion. AI-specific runtime threats can appear in model inputs, outputs and tool interactions; model scanning prior to release cannot inspect every later conversation.