GDPR Recital 26 - definition
General Data Protection Regulation (GDPR) Recital 26 is the EU interpretive text that explains where personal data ends and anonymous information begins. It belongs to Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016, which has applied since 25 May 2018. Recital 26 supports Article 4(1) GDPR, which defines personal data as information relating to an identified or identifiable natural person.
The key test in Recital 26 is whether a person is identifiable by means that are reasonably likely to be used. The recital states that identification must consider “all the means reasonably likely to be used” by the controller or by another person. It also requires objective factors, including cost, time, available technology at the time of processing, and technological development, to be assessed.
For image and video anonymization, Recital 26 means that a blurred face or blurred license plate is not automatically anonymous. The result is anonymous only if the remaining footage does not allow a person to be identified with reasonable means. If the original unblurred video is retained, that original file remains personal data. If the blurred copy can still be linked to a person through voice, clothing, location, time, metadata, distinctive behavior, or other context, the blurred copy may also remain personal data.
Role of GDPR Recital 26 in video anonymization
Recital 26 is central when deciding whether GDPR continues to apply to processed video. It separates anonymization from weaker privacy measures such as pseudonymization, access control, encryption, or partial masking.
Processing result | GDPR status | Example in footage |
|---|---|---|
Identifiable footage | Personal data under Article 4(1) GDPR | A face or license plate is clear enough to identify a person or a natural person associated with the vehicle. |
Pseudonymized footage | Still personal data where attribution remains possible; Article 4(5) GDPR defines pseudonymization | Faces are replaced with codes, but a separate key can restore identity. |
Anonymized footage | Outside GDPR, according to Recital 26 | Faces and license plates are irreversibly masked, and no other reasonable identifier remains. |
In practice, the assessment is contextual. Closed-circuit television footage from a private workplace may require stricter review than a low-resolution public crowd scene, because colleagues, managers, or regular visitors may identify a person from context even when the face is blurred.
Reasonable likelihood test for faces and license plates
The Recital 26 test is not a mathematical threshold set by the GDPR. It is a risk-based legal assessment supported by technical evidence. The Court of Justice of the European Union used a similar approach in Case C-582/14, Breyer, judgment of 19 October 2016, where identifiability depended on whether lawful and practical means of identification were reasonably available.
A practical assessment for image and video files should record at least these factors:
- Identifier strength: whether faces, license plates, voices, tattoos, uniforms, location signs, or metadata can identify a person.
- Adversary model: whether identification is likely by the publisher, recipient, data subject, employer, authority, or public viewer.
- Technical reversibility: whether blur, pixelation, compression, or masking can be reversed or bypassed.
- Contextual uniqueness: whether the scene, event time, location, or behavior narrows the possible individuals.
- Residual linkage: whether the anonymized copy can be linked to the original file, access logs, filenames, or case records.
A simple working formula is:
Anonymization under Recital 26 = irreversible transformation + no reasonably likely re-identification means + documented context assessment
Technical implications for face blurring and license plate blurring
Face detection and license plate detection are technical steps that support anonymization, but they do not decide the legal result. The legal result depends on whether the processed output meets the Recital 26 standard.
For compliance review, organizations should measure and document detection and masking quality. Relevant operational parameters include false negatives, false positives, detection confidence thresholds, manual review rate, frame coverage, and output resolution. In video, frame-by-frame consistency matters because one missed frame can preserve an identifying face or plate.
Parameter | Why it matters under Recital 26 |
|---|---|
False negative rate | Undetected faces or license plates may leave personal data in the published file. |
Mask permanence | Irreversible masking reduces the chance that the image can be reconstructed. |
Manual quality review | Review helps detect missed identifiers such as badges, monitors, documents, tattoos, or logos. |
Retention of originals | Original unmasked footage remains personal data and requires a separate GDPR basis and security controls. |
Gallio PRO automatically detects and blurs faces and license plates in recorded images and videos. It does not perform real-time anonymization or video stream anonymization. It does not automatically detect company logos, tattoos, name badges, documents, or content shown on monitors, but these elements can be masked manually with the built-in editor. Gallio PRO does not collect logs containing face or license plate detections, and it does not collect logs containing personal data or special categories of personal data.
EU legal references for GDPR Recital 26
Recital 26 should be read together with GDPR provisions and European supervisory guidance. The most relevant sources for video anonymization are listed below.
Source | Relevance |
|---|---|
GDPR Article 4(1) | Defines personal data and identifiable natural person. |
GDPR Article 4(5) | Defines pseudonymization. |
GDPR Article 5(1)(c) | Requires data minimization. |
GDPR Article 25 | Requires data protection by design and by default. |
GDPR Article 32 | Requires security appropriate to the risk. |
European Data Protection Board Guidelines 3/2019 on processing of personal data through video devices, version 2.0, adopted 29 January 2020 | Explains GDPR requirements for video surveillance and the treatment of identifiable images. |
Article 29 Working Party Opinion 05/2014 on Anonymisation Techniques, WP216, adopted 10 April 2014 | Describes anonymization risks, including singling out, linkability, and inference. |
US equivalent
There is no single US equivalent to GDPR Recital 26. The closest concepts appear in specific frameworks, including deidentified information under the California Consumer Privacy Act as amended by the California Privacy Rights Act, California Civil Code section 1798.140(m), and de-identification of protected health information under the Health Insurance Portability and Accountability Act regulations at 45 CFR 164.514(b). These US rules are sectoral or state-specific, while Recital 26 is part of an EU-wide personal data framework.
See also
The following related glossary entries provide additional context for image and video anonymization workflows.