NCA-GENL Trustworthy AI Practice Question
A media company uses an LLM to generate article drafts. Legal requires that the system never reproduce long verbatim passages from copyrighted training sources. Which mitigation most directly reduces this risk at generation time?
⚠ Common exam trap
The trap here is assuming a system prompt that says 'paraphrase' reliably prevents verbatim copying, when only an output-side overlap check actually detects it.
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
✓
Apply a decoding-time constraint that blocks or rewrites outputs containing long n-gram matches against a reference corpus of copyrighted text.
Copyright risk from verbatim reproduction is best addressed by detecting long overlapping sequences between generated output and known source text, then blocking or rewriting those spans before delivery. This is a generation-time control that directly measures the prohibited behavior. Temperature changes, prompt instructions, and stylistic fine-tuning do not verify overlap and therefore cannot guarantee that infringing passages are stopped.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Fine-tune the model on a corpus of original company articles so that its outputs resemble proprietary writing rather than external sources.
Why it's wrong here
Fine-tuning shifts style toward the company corpus but does not remove memorized passages from the base model, and it provides no runtime detection of verbatim overlap. The model can still reproduce copyrighted text when prompted or when such text is high-probability. This approach changes tone without addressing the specific legal risk of long verbatim reproduction.
- ✗
Lower the sampling temperature to make the model more deterministic and consistent in its phrasing.
Why it's wrong here
Lower temperature reduces randomness, which can actually make the model more likely to emit high-probability sequences that may correspond to memorized training text. It does not compare output against copyrighted sources and cannot detect or prevent verbatim overlap. Determinism is unrelated to the legal concern, so this change may worsen rather than mitigate the risk.
- ✗
Add a system prompt instructing the model to always paraphrase and never quote more than a few consecutive words from any source.
Why it's wrong here
Prompt instructions influence style but are not a reliable enforcement mechanism; models can ignore them, especially under adversarial prompting or when a memorized passage is strongly favored. There is no verification that the instruction was followed, so verbatim spans can still reach the user. A prompt is a soft control, while the legal requirement demands a check that actually detects overlap.
- ✓
Apply a decoding-time constraint that blocks or rewrites outputs containing long n-gram matches against a reference corpus of copyrighted text.
Why this is correct
Verbatim reproduction is detectable as unusually long overlapping n-grams between output and source text, so checking generated spans against a reference corpus at decoding time directly targets the risk. Blocking or rewriting flagged spans prevents the infringing text from being delivered. This operates at generation time, matching the legal requirement, and does not depend on the model having memorized less during training.
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.