UK GDPR Recital 26 is the test that decides whether a dataset has actually left personal-data territory: data is anonymous only once no one can reasonably re-identify a person from it, by any means reasonably likely to be used. The ICO's own guidance applies that test through its motivated-intruder standard — asking whether a reasonably competent person, motivated to try, could re-link the data using publicly available information. A rare combination of amount, date, and dispute reason can fail that test even with names removed. PCI DSS v4.0 still limits the account fields on top. anonym.plus marks each value on your device, so the set stays analysable while customer data is shielded.
When this applies
A dispute file holds thousands of rows with account, customer, and reason columns. You strip those identifiers under the standard before any model is trained.
How anonym.plus handles it
- Open the file in anonym.plus on your device.
- The tool maps account, customer, and contact columns.
- It flags each sensitive cell across the rows.
- Keep reason codes and amount columns for analysis.
- Turn off the name map for true anonymity.
- Save the anonymised set locally.
What you need to provide
- The dataset (CSV, JSON, or Parquet-as-CSV).
- An operator (Replace keeps columns analysable).
- The name map turned OFF for non-reversible output.
PII & financial identifiers detected
| Category | anonym.plus entity type | Example |
|---|---|---|
| Account | CREDIT_CARD | pan column → [CARD] |
| Names | PERSON | customer column → [BUYER] |
| Contact | EMAIL_ADDRESS | email column → [EMAIL] |
| Location | LOCATION | billing town → [TOWN] |
| Amount | MONEY | dispute amount → kept |
| Dates | DATE_TIME | dispute date → [DATE] |
Compliance achieved
- Limits stored account data per PCI DSS v4.0.
- Works toward the anonymity threshold in UK GDPR Recital 26: no one reasonably able to re-identify a person from the result.
- Applies the ICO's motivated-intruder test as the practical check for whether a rare combination of fields could still single someone out.
- Keeps reason and amount columns for analysis.
- Offline handling keeps the dataset off any server.
Anonymise dispute datasets offline — see plans & start free →
Limitations & cautions
A rare combination of amount, date, and town can still single out a customer even with no name. Review quasi-identifiers, and keep the name map off when you need true anonymity.
Frequently asked questions
What does the motivated-intruder test actually ask?
Whether a reasonably competent person, who wanted to identify someone in the dataset, could do so using the data itself plus other information already publicly available. It is the ICO's practical way of applying the Recital 26 anonymity standard rather than leaving it abstract.
Can I keep the reason and amount columns?
Yes. Those stay so analysis and training still work; only personal cells are removed. But review whether a rare combination of amount, date, and reason could still pass the motivated-intruder test in reverse.
Does the dataset get uploaded during anonymisation?
No. The app is fully offline, so the data never leaves your machine.