A record of datasets: Which artifact most directly satisfies this need?

Documenting datasets, splits, hyperparameters, and metrics is exactly what lets an auditor validate and reproduce training and testing results.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. A signed acceptable-use policy defining who may operate the system and what prohibited uses apply once it is in production.

   Plausible but wrong: an acceptable-use policy governs deployment behavior, not the reproducibility of training and testing results.
2. A vendor licensing summary listing the model's usage rights, indemnities, and the supplier's stated performance benchmarks.

   Almost right because it references benchmarks, but licensing records describe an acquired product rather than the team's own test procedure.
3. A stakeholder communications plan describing how evaluation outcomes will be disclosed to affected users and regulators.

   Plausible but wrong: a communications plan supports transparency disclosures, not the reproduction and validation of the metrics themselves.
4. A record of datasets, splits, hyperparameters, and evaluation metrics captured to validate results and support compliance. ✓

   Correct because III.B.5 documents the training/testing process specifically to validate results, establish compliance, and manage risks.

How to remember it

Documenting datasets, splits, hyperparameters, and metrics is exactly what lets an auditor validate and reproduce training and testing results.

More Understanding How to Govern AI Development questions

• Document the proxy-bias and overfitting findings: Which action best fits a training-and-testing →
• A model card summarizing intended use: Which readiness artifact directly addresses the standardized summary →
• Track performance against thresholds continuously: Which response best reflects that maintenance obligation? →
• All 426 Understanding How to Govern AI Development questions →