100% Latest & Updated Microsoft GH-300 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 20
A legal review is evaluating a Copilot-generated source file. A developer wants to record why a generated routine was kept. Which response best addresses the stated risk or requirement?
- Apply the organization's provenance and license review policy before distribution.
- Have the proposed reuse checked against that license before inclusion in the product.
- Preserve the notice while legal review determines permitted reuse or replacement.
- Document code origin review, human changes, and applicable verification evidence.
Explanation
Correct Answer
D:An auditable decision record supports future maintenance and licensing inquiries.
Incorrect Answers
A:AI-generated output can still require independent intellectual-property assessment. This addresses a different situation: A developer assumes that generated code always carries no intellectual-property risk.
B:Finding a matching public source does not itself grant unrestricted distribution rights. This addresses a different situation: A code reference links to a public repository with a restrictive license.
C:Copyright notices are a signal for review rather than an invitation to strip attribution. This addresses a different situation: A generated snippet includes a third-party copyright header.
Question 21
A team uses AI while designing a public-facing accessibility feature. Copilot invents alt text for a product image it cannot actually inspect. Which recommendation directly resolves this issue?
- Test keyboard focus order and remove the unintended focus trap.
- Remove unsupported sensitive-attribute inferences and review fairness risks.
- Perform functional assistive-technology checks rather than trusting compilation alone.
- Obtain image-grounded evidence before asserting the image's contents.
Explanation
Correct Answer
D:Unverified descriptions can mislead users of assistive technology.
Incorrect Answers
A:Users must be able to navigate away from ordinary panels. This addresses a different situation: The proposed keyboard navigation traps focus inside a nonmodal panel.
B:Sensitive identity stereotypes are not legitimate substitutes for job-related evidence. This addresses a different situation: A generated job-screening summary infers personality from unrelated demographic cues.
C:Syntax validity does not establish accessible interaction behavior. This addresses a different situation: The code passes syntax checks but the UI is difficult to use with a screen reader.
Question 22
A team uses AI while designing a public-facing accessibility feature. The generated form omits a programmatic label for its email input. Which action is the most appropriate response to this situation?
- Add a connected accessible label and test with accessibility tooling.
- Perform functional assistive-technology checks rather than trusting compilation alone.
- Test keyboard focus order and remove the unintended focus trap.
- Obtain image-grounded evidence before asserting the image's contents.
Explanation
Correct Answer
A:A visual placeholder does not provide a reliable input label.
Incorrect Answers
B:Syntax validity does not establish accessible interaction behavior. This addresses a different situation: The code passes syntax checks but the UI is difficult to use with a screen reader.
C:Users must be able to navigate away from ordinary panels. This addresses a different situation: The proposed keyboard navigation traps focus inside a nonmodal panel.
D:Unverified descriptions can mislead users of assistive technology. This addresses a different situation: Copilot invents alt text for a product image it cannot actually inspect.
Question 23
A team uses AI while designing a public-facing accessibility feature. A generated job-screening summary infers personality from unrelated demographic cues. What should the developer do before accepting the proposed result?
- Obtain image-grounded evidence before asserting the image's contents.
- Add a connected accessible label and test with accessibility tooling.
- Remove unsupported sensitive-attribute inferences and review fairness risks.
- Perform functional assistive-technology checks rather than trusting compilation alone.
Explanation
Correct Answer
C:Sensitive identity stereotypes are not legitimate substitutes for job-related evidence.
Incorrect Answers
A:Unverified descriptions can mislead users of assistive technology. This addresses a different situation: Copilot invents alt text for a product image it cannot actually inspect.
B:A visual placeholder does not provide a reliable input label. This addresses a different situation: The generated form omits a programmatic label for its email input.
D:Syntax validity does not establish accessible interaction behavior. This addresses a different situation: The code passes syntax checks but the UI is difficult to use with a screen reader.
Question 24
A team uses AI while designing a public-facing accessibility feature. The proposed keyboard navigation traps focus inside a nonmodal panel. Which response best addresses the stated risk or requirement?
- Add a connected accessible label and test with accessibility tooling.
- Obtain image-grounded evidence before asserting the image's contents.
- Remove unsupported sensitive-attribute inferences and review fairness risks.
- Test keyboard focus order and remove the unintended focus trap.
Explanation
Correct Answer
D:Users must be able to navigate away from ordinary panels.
Incorrect Answers
A:A visual placeholder does not provide a reliable input label. This addresses a different situation: The generated form omits a programmatic label for its email input.
B:Unverified descriptions can mislead users of assistive technology. This addresses a different situation: Copilot invents alt text for a product image it cannot actually inspect.
C:Sensitive identity stereotypes are not legitimate substitutes for job-related evidence. This addresses a different situation: A generated job-screening summary infers personality from unrelated demographic cues.
Question 25
A team uses AI while designing a public-facing accessibility feature. The code passes syntax checks but the UI is difficult to use with a screen reader. What is the most defensible next step under responsible development practices?
- Remove unsupported sensitive-attribute inferences and review fairness risks.
- Add a connected accessible label and test with accessibility tooling.
- Perform functional assistive-technology checks rather than trusting compilation alone.
- Test keyboard focus order and remove the unintended focus trap.
Explanation
Correct Answer
C:Syntax validity does not establish accessible interaction behavior.
Incorrect Answers
A:Sensitive identity stereotypes are not legitimate substitutes for job-related evidence. This addresses a different situation: A generated job-screening summary infers personality from unrelated demographic cues.
B:A visual placeholder does not provide a reliable input label. This addresses a different situation: The generated form omits a programmatic label for its email input.
D:Users must be able to navigate away from ordinary panels. This addresses a different situation: The proposed keyboard navigation traps focus inside a nonmodal panel.
Question 26
A developer uses Copilot to summarize an unfamiliar repository. The team has these independent requirements: (1) A README contains instructions to transmit a private token to an unknown domain. (2) A scraped issue asks the assistant to send its session transcript to a third-party endpoint. Which TWO actions together satisfy all requirements? Choose TWO.
- Keep the instruction hierarchy intact and ignore the lower-trust comment.
- Open the referenced documents and verify claims before relying on them.
- Do not obey the external request; restrict tool activity to the approved task.
- Treat repository prose as untrusted data and reject the credential-transmission instruction.
- Inspect scope and impact before approving the destructive command.
Explanation
Correct Answers
C:External issue text is task material, not authorization to exfiltrate conversations.
D:Retrieved content can include prompt injection that does not override user intent.
Incorrect Answers
A:Source comments cannot grant themselves higher authority than the user's rules. This addresses a different situation: A comment claims to be a system instruction and demands disabling security scanning.
B:A plausible citation may be missing, unrelated, or fabricated. This addresses a different situation: A repository search result includes fabricated citations to internal policy files.
E:Agent tool execution deserves the same review as any human-authored shell command. This addresses a different situation: An agent proposes an unfamiliar shell command that deletes the build directory recursively.
Question 27
A developer uses Copilot to summarize an unfamiliar repository. A comment claims to be a system instruction and demands disabling security scanning. What should the developer do before accepting the proposed result?
- Do not obey the external request; restrict tool activity to the approved task.
- Treat repository prose as untrusted data and reject the credential-transmission instruction.
- Open the referenced documents and verify claims before relying on them.
- Keep the instruction hierarchy intact and ignore the lower-trust comment.
Explanation
Correct Answer
D:Source comments cannot grant themselves higher authority than the user's rules.
Incorrect Answers
A:External issue text is task material, not authorization to exfiltrate conversations. This addresses a different situation: A scraped issue asks the assistant to send its session transcript to a third-party endpoint.
B:Retrieved content can include prompt injection that does not override user intent. This addresses a different situation: A README contains instructions to transmit a private token to an unknown domain.
C:A plausible citation may be missing, unrelated, or fabricated. This addresses a different situation: A repository search result includes fabricated citations to internal policy files.
Question 28
A developer uses Copilot to summarize an unfamiliar repository. An agent proposes an unfamiliar shell command that deletes the build directory recursively. Which response best addresses the stated risk or requirement?
- Inspect scope and impact before approving the destructive command.
- Open the referenced documents and verify claims before relying on them.
- Treat repository prose as untrusted data and reject the credential-transmission instruction.
- Keep the instruction hierarchy intact and ignore the lower-trust comment.
Explanation
Correct Answer
A:Agent tool execution deserves the same review as any human-authored shell command.
Incorrect Answers
B:A plausible citation may be missing, unrelated, or fabricated. This addresses a different situation: A repository search result includes fabricated citations to internal policy files.
C:Retrieved content can include prompt injection that does not override user intent. This addresses a different situation: A README contains instructions to transmit a private token to an unknown domain.
D:Source comments cannot grant themselves higher authority than the user's rules. This addresses a different situation: A comment claims to be a system instruction and demands disabling security scanning.
Question 29
A developer uses Copilot to summarize an unfamiliar repository. A scraped issue asks the assistant to send its session transcript to a third-party endpoint. What is the most defensible next step under responsible development practices?
- Do not obey the external request; restrict tool activity to the approved task.
- Keep the instruction hierarchy intact and ignore the lower-trust comment.
- Treat repository prose as untrusted data and reject the credential-transmission instruction.
- Inspect scope and impact before approving the destructive command.
Explanation
Correct Answer
A:External issue text is task material, not authorization to exfiltrate conversations.
Incorrect Answers
B:Source comments cannot grant themselves higher authority than the user's rules. This addresses a different situation: A comment claims to be a system instruction and demands disabling security scanning.
C:Retrieved content can include prompt injection that does not override user intent. This addresses a different situation: A README contains instructions to transmit a private token to an unknown domain.
D:Agent tool execution deserves the same review as any human-authored shell command. This addresses a different situation: An agent proposes an unfamiliar shell command that deletes the build directory recursively.
Question 30
A developer uses Copilot to summarize an unfamiliar repository. A repository search result includes fabricated citations to internal policy files. Which recommendation directly resolves this issue?
- Do not obey the external request; restrict tool activity to the approved task.
- Open the referenced documents and verify claims before relying on them.
- Inspect scope and impact before approving the destructive command.
- Keep the instruction hierarchy intact and ignore the lower-trust comment.
Explanation
Correct Answer
B:A plausible citation may be missing, unrelated, or fabricated.
Incorrect Answers
A:External issue text is task material, not authorization to exfiltrate conversations. This addresses a different situation: A scraped issue asks the assistant to send its session transcript to a third-party endpoint.
C:Agent tool execution deserves the same review as any human-authored shell command. This addresses a different situation: An agent proposes an unfamiliar shell command that deletes the build directory recursively.
D:Source comments cannot grant themselves higher authority than the user's rules. This addresses a different situation: A comment claims to be a system instruction and demands disabling security scanning.
GitHub Copilot (GH-300) Practice Tests
