How do we test generated case records for disclosure risk?

**No test can prove the absence of risk, but several checks are worth running.**

**Near-copy search.** If you hold any real or third-party records in the same domain, compare generated text and field combinations against them. Close matches in rationales or unusual amounts deserve a closer look.

**Rare combinations.** List attribute combinations that occur only once, such as an uncommon product, role and amount together. Rare combinations are where a generated record could resemble something specific.

**Free-text review.** Read a sample of handover notes and rationales for oddly specific detail, such as unusual place names, dates or phrasing that seems lifted from a source.

**Identifier formats.** Confirm identifiers like `APP-2187` follow the generated pattern and do not resemble real account or tax number formats.

**Generation notes.** Ask how the release was produced, what grounding it reflects, and what was done to limit leakage.

**What to say in your listing.** State that the data is generated and that buyers should run their own disclosure checks. Do not describe the release as anonymous, privacy-compliant or approved by any regulator.

If a check raises a concern, report the specific records back and ask for a corrected release before you list it.