Question 890
At an enterprise inventorying information for a new model, an AI project leader faces this situation: A safety classifier has almost no representative hazardous events but leaders request production approval anyway. Available records are not automatically relevant or authorized for training. What aspect of AI project delivery best explains the observation?
Explanation
Correct Answer
In an enterprise inventorying information for a new model, the underlying issue is: A go/no-go decision needs adequate examples for critical outcomes and known performance uncertainty. It explains the scenario: A safety classifier has almost no representative hazardous events but leaders request production approval anyway. The appropriate project response is: Require a rare-event sufficiency assessment and choose a controlled pilot or defer deployment. This can be checked by: The gate record states whether error uncertainty meets the approved safety tolerance.
Incorrect Answers
Incremental extraction can reduce cost when reliable change-tracking boundaries are available. The appropriate correction for the reported case is: Require a rare-event sufficiency assessment and choose a controlled pilot or defer deployment.
Dataset availability and technical access are separate from contractual use and model-training permissions. Fixing that different mechanism would not address the current root cause: a go/no-go decision needs adequate examples for critical outcomes and known performance uncertainty.
Raw record count alone does not establish that target classes have enough useful learning examples. Its verification is distinct from the reported condition: A safety classifier has almost no representative hazardous events but leaders request production approval anyway. Available records are not automatically relevant or authorized for training.