Validate captions on representative data and test leakage: Which validation plan best addresses these

Quality validation and security testing answer different questions; both are required before deployment and during monitoring.

The question

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

1. Validate captions on representative data and test leakage and security. ✓

   Representative quality testing and separate leakage and security testing provide evidence for both stated acceptance requirements.
2. Red-team prompt leakage against the production model.

   Adversarial leakage testing is useful, but it does not validate caption quality on representative footage.
3. Benchmark quality only, then rely on user reports for disclosure incidents.

   A quality benchmark is necessary but relying on post-deployment reports leaves disclosure risks insufficiently tested before launch.
4. Check access permissions for footage and prompts.

   Access-control review addresses one security measure but does not establish caption quality or test model-mediated leakage.

How to remember it

Quality validation and security testing answer different questions; both are required before deployment and during monitoring.

More Understanding How to Govern AI Development questions

• Use a cost-sensitive threshold: Which approach best addresses both the asymmetric error costs and limited →
• Test synthetic records for privacy leakage before use: Which action is most immediately necessary? →
• Adjudicate sampled disagreements and revise the labeling: What should the team do first? →
• All 426 Understanding How to Govern AI Development questions →