Personal Data Anonymization in the Age of Powerful AI Models

Personal Data Anonymization in the Age of Powerful AI Models

Published
November 19, 2026
Author
Khrabrykh S. A.
Original language
English

The report examines the anonymisation of personal data at a time when AI models can match heterogeneous attributes, infer a person’s likely characteristics and reproduce fragments of their training material. The study addresses four questions: when information can be regarded as anonymous in the legal sense; how to assess the residual risks of singling out, linkage and inference; what evidence a developer or dataset owner should provide; and how the assessment changes when moving from a de-identified table to a trained model, a vector index, a synthetic dataset or published statistics.

The legal basis of the analysis comprises the GDPR, the case law of the Court of Justice of the EU in Breyer, IAB Europe and EDPS v SRB, EDPB Opinion 28/2024 on AI models, and the draft Guidelines 02/2026 on Anonymisation with its No Record Isolation, No Linkage and No Inference criteria; at the date of the report the draft remains under public consultation. The report reviews k-anonymity, l-diversity and t-closeness, differential privacy and privacy budget accounting, synthetic data, PETs and federated learning, membership inference attacks, model inversion and regurgitation, as well as the comparability of the EU, UK and US approaches.

The main conclusion is that anonymity is a contextual, verifiable and revisable assessment, not the name of a procedure for removing identifiers. No single technique replaces an examination of the means of identification reasonably likely to be used; a model trained on personal data does not automatically become anonymous, and embeddings and retrieval systems require their own testing. In sensitive fields — international searches for wanted persons, extradition, sanctions and AML/KYC compliance, migration — a unique sequence of procedural events can identify a person faster than a name, and an error risks falsely attributing criminal or sanctions status.

Observatoire ARGA proposes a six-stage methodology for independent anonymity audits, an evidentiary anonymisation passport, testing protocols, a pilot programme based on artificial data and draft comments on the EDPB document. A decision against open publication is treated as a fully valid professional outcome rather than a failure of the research.

The report is normative and methodological in nature: it contains no results from testing real systems or wanted-person registers, its numerical examples are illustrative, and the absence of a successful attack in an experiment does not prove that future extraction of information is impossible. Internal templates, the authors stress, should not be called a “GDPR certificate”.

Full text of the document
ARGA Observatory

Citation Rules

  1. Mandatory source attributionWhen using ARGA Observatory materials, the full name must be cited.
  2. Date and version indicationFor analytical reports, the year of publication is mandatory.
  3. Link to the originalElectronic materials must be accompanied by an active link to the official website.
  4. Context preservationCitations must not be shortened or altered in a way that distorts the original meaning.
  5. Note on adaptationIf the text is abridged or translated, state: "adapted from ARGA Observatory report".
  6. No commercial use without written permission from the organization
  7. Data accuracy preservationCharts and tables must be reproduced without changes.
Confidential enquiry

Ask the ARGA experts

Four short steps. We reply within one business day.

Step 1 of 4
What situation are you facing?
What stage is the case at?
Describe the situation
You can skip this field.
How should we reach you?
Your enquiry goes straight to ARGA experts. We do not publish it or pass it to third parties.
Official registration
ARGA in international registries
UNGM ID
1232417
ARGA is registered in the United Nations procurement system.
View registration
UN Global Compact
213173
Participant of the UN Global Compact.
ADB GMS Number
049963
Asian Development Bank — registered participant.
Scroll to Top