The correct answer is The filter evaluates proposed suggestions against a large set of well known open source repositories and flags a suggestion when a similarity threshold is met.

This reflects how duplication detection filters typically work because they compare a proposed completion to a corpus of public open source code and compute similarity. When the match surpasses a configured threshold the assistant warns or filters the suggestion. The limitation is that this approach does not guarantee that every accepted suggestion is entirely novel and it does not automatically rewrite content into a unique form. It may also allow short or common snippets to pass since these are often excluded to reduce noise and it generally does not check against your private code unless a separate mechanism exists.

The filter indexes your organization's projects in Cloud Source Repositories and blocks any suggestion that matches internal code across those private repos is incorrect because these filters are designed to compare against public code rather than indexing your internal repositories. In addition Cloud Source Repositories has been deprecated which makes this option even less likely to appear on newer exams.

The filter only flags code that appears frequently in public repositories and it automatically rewrites those completions into unique alternatives is incorrect because the filter relies on similarity thresholds rather than popularity or frequency and it flags or blocks rather than rewriting suggestions automatically.

The filter performs exact whole file matching of your workspace and warns you only when an entire file would be duplicated is incorrect because duplication checks target segments of code and use similarity scoring. Whole file matching would be impractical and would miss the intended goal of discouraging verbatim reuse of smaller blocks.

Full AWS Practitioner Certification Question