The correct option is Reducing the volume to be examined by reviewing representative subsets.

Sampling reduces the number of records auditors must inspect while allowing them to draw conclusions about the entire population when the sample is representative. A statistical or judgmental sample gives reasonable assurance with far less work than a full review and that makes validation feasible in very large datasets.

Reviewing every record to ensure complete validation is incorrect because inspecting every record is often impractical or impossible when datasets are huge and that is why auditors use sampling to obtain reasonable assurance efficiently.

Relying on Cloud Audit Logs as a substitute for sampling is incorrect because logs are a source of evidence that can support testing but they do not replace the need to select representative records for validation. Logs may complement sampling but they do not themselves perform statistical inference about a population.

Increasing the probability of analyst bias during validation is incorrect because properly designed sampling methods reduce bias and increase objectivity. Poor sample design can introduce bias but the purpose of sampling is to make validation systematic and defensible rather than to increase analyst bias.

Full AWS Practitioner Certification Question