NCA-GENL Trustworthy AI Practice Question
A media company uses an LLM to generate summaries of user-submitted articles. Legal counsel requires that the system detect and refuse requests that attempt to extract verbatim copyrighted passages longer than a defined threshold. Which capability should the team implement?
⚠ Common exam trap
The trap here is choosing input-side or training-side measures for a risk that only manifests in the generated output.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Output filtering that compares generated text against source documents and blocks responses exceeding the verbatim length threshold.
The requirement is a measurable output constraint: no verbatim spans above a defined length. Only output filtering that compares generated text against the source and blocks violations enforces that threshold deterministically. Input keyword blocking, larger context windows, and stylistic fine-tuning may influence behavior but cannot guarantee or verify the specific legal limit, leaving the organization exposed.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Output filtering that compares generated text against source documents and blocks responses exceeding the verbatim length threshold.
Why this is correct
This directly implements the legal requirement by inspecting generated output for verbatim overlap with source material and refusing responses that exceed the configured length. It is a concrete, testable control that operates at the point of release, ensuring the system never returns the prohibited passages. It also produces an auditable record of blocked responses for compliance review.
- ✗
Input filtering that rejects any prompt containing words like summarize or excerpt.
Why it's wrong here
Blocking prompts based on keywords is brittle and misaligned with the actual risk, which is verbatim reproduction in the output. Legitimate summarization requests would be rejected while paraphrased extraction attempts would pass. This approach does not measure overlap with source documents, so it cannot enforce the length threshold legal counsel specified.
- ✗
Fine-tuning the model on public-domain summaries so it learns to paraphrase.
Why it's wrong here
Fine-tuning can shift style toward paraphrasing, but it offers no guarantee or measurable threshold for verbatim length. The model could still occasionally reproduce long exact passages, which is exactly what legal counsel wants prevented. Without a runtime check against the source, there is no way to demonstrate compliance or catch failures before release.
- ✗
Increasing the model's context window so it can consider the entire source document when summarizing.
Why it's wrong here
A larger context window may improve summary fidelity, but it does not prevent the model from emitting long verbatim spans. In fact, more complete source context can make exact reproduction easier. This option addresses input capacity rather than output compliance and leaves the copyright exposure entirely unmitigated.
About these practice questions
Courseiva writes every NCA-GENL question from scratch — 367 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official NVIDIA exam blueprint
This NCA-GENL practice question is part of Courseiva's free NVIDIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the NCA-GENL exam.