An organization wants to create an AI audit trail that satisfies regulatory requirements. Which data should be captured at minimum in the audit log for each AI decision?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you need to capture the input data used, the model version, the output produced, the timestamp, and whether a human or AI made the decision. That gives you traceability, reproducibility, and accountability for each decision.
Full Explanation
The correct minimum audit log for each AI decision should capture the input data used, the model version, the output produced, the timestamp, and the decision-maker, whether human or AI. These elements let reviewers trace what information drove the decision, reproduce the decision with the same model version, verify what was produced, establish when it happened, and assign accountability. The model's source code, training dataset, and hyperparameter configuration are important documentation, but they are not per-decision log requirements. The full probability distribution across all possible model outputs for each decision is useful for some analyses, but it is not a minimum requirement. The identities of all data scientists involved in building and deploying the model belong in governance records, not per-decision logs.