Guide · Data & pseudonymisation
Anonymisation and pseudonymisation: what is the difference?
Anonymisation aims to make identifying a person impossible in practice and irreversibly. Pseudonymisation replaces identifying information with aliases or identifiers, while keeping separately the elements that allow re-attribution. Pseudonymised data therefore remains, in principle, personal data.
Two different goals
The two terms are often used as synonyms, especially when a document is “anonymised” by hiding names. In data protection law, however, the distinction is essential. Effective anonymisation must prevent, in practice, any identification of the person by means reasonably likely to be used. Pseudonymisation, by contrast, reduces the direct link with identity without seeking to remove it permanently.
| Pseudonymisation | Anonymisation | |
|---|---|---|
| Principle | Replace or separate identifiers and keep separately the information that allows re-attribution. | Make the person unidentifiable, effectively and irreversibly. |
| Reversibility | Yes, when the additional information is available. | No, if the anonymisation is truly effective. |
| GDPR | The data remains personal data when re-identification is still possible. | Effectively anonymous data no longer falls under the GDPR as personal data. |
| Practical value | Keep the document useful while reducing exposure of identity. | Share or use data without keeping a reasonable link with the people concerned. |
Removing a name is not always enough
A document can remain identifiable even when first and last names have gone. An address, a unique number, a very specific job title, a rare date or a combination of several items can lead back to the person concerned. You must therefore think about the overall risk of re-identification, not only about whether a name is present.
Replacing “ Claire Martin ” with “ PERS_1 ” reduces direct identification. But if the document states that she is the only director of a given establishment, with a precise date and place, her identity may still be easy to infer. The document is then pseudonymised, not necessarily anonymous.
Why pseudonymisation is useful with AI
For many professional uses, the goal is not to destroy the link with people permanently: you need a summary, a rewording or an analysis, then to put the result back in context. Pseudonymisation keeps a consistent structure — for example PERS_1 and PERS_2 for two people — while avoiding sending certain identifiers directly to the AI provider when they are not needed for the task.
This approach does not turn the data into anonymous information: it reduces exposure without removing the obligations that still apply to the processing.
What this changes under the GDPR
Pseudonymisation is a protective measure encouraged by the GDPR, but it does not remove the other obligations. If data remains re-attributable, you still need an appropriate legal basis, to limit data to what is necessary, define its retention, secure access and respect people’s rights where they apply. Effective anonymisation follows a different logic: truly anonymous data is no longer personal data.
How NYMETRA approaches this problem
NYMETRA was designed around pseudonymisation and user control. The software detects, in the client’s environment, the information that may need protecting, then lets the user check and correct the selection. Approved items are replaced before being sent to the chosen AI provider. The mapping table stays in the client’s environment and, where the mapping allows, makes it possible to restore the original information locally in the answer.
This process is deliberately described as pseudonymisation: NYMETRA does not present a reversible process as irreversible anonymisation.
If you need to be able to recover the original information after processing, you are generally dealing with pseudonymisation. If the link with the person must be removed permanently, the anonymisation requirement is much stricter.