Visual data anonymization means transforming images or videos so that individuals cannot be identified by any means reasonably likely to be used. Under GDPR and UK GDPR, data is no longer personal when identification is not possible, taking into account costs, time, and available technology [1]. In publishing scenarios, anonymization typically relies on face blurring and license plate blurring, but many other visual cues can still identify a person. This article focuses on what else commonly requires attention before publishing photos or videos.
Plates are only one of several common identifiers. In practice, people are often recognised by facial features, unique clothing, tattoos, workplace badges, or the combination of location, time, and context. Regulators regularly note that identifiability is not limited to names or numbers. It includes any information relating to an identified or identifiable natural person, including by reference to physical, physiological, genetic, mental, economic, cultural, or social identity [1]. A robust approach looks beyond vehicles to the broader visual scene.
This checklist lists typical visual elements that can reveal identity. The items are ordered from most to least likely to enable identification in common publishing use cases. Organisations often consider these steps as part of a common compliance approach when preparing media for websites, reports, social channels, or public information portals.
Public places do not create a free-to-publish zone. If a person is identifiable, GDPR or UK GDPR applies. A contextual risk assessment is still needed before posting.
There are narrow cases where anonymization may not be required. These do not remove the need for assessment and safeguards:
Law enforcement processing by competent authorities is governed by a separate regime (in the EU: Law Enforcement Directive as implemented in national law; in the UK: DPA 2018 Part 3) and is not covered here. For business and public-sector publishing, the above three are the most relevant boundary cases.
Automation helps scale visual data anonymization. Tools can detect faces and license plates, then apply masking or blurring. Accuracy and cost depend on scene complexity, camera angles, lighting, and model quality. A human review step is still common practice for high-risk publications or where sensitive contexts appear.
On-premise software is often preferred where footage is sensitive or where transfer to third-party servers is restricted by policy. Keeping processing on controlled infrastructure can reduce data transfer and vendor exposure risk. Where cloud is used, publishers typically consider data location, retention, access controls, and logging.
Check out Gallio PRO for automated face blurring and license plate blurring with deployment options aligned to enterprise governance.
Teams seeking a hands-on evaluation can Download a demo to test workloads on representative footage.
Topic | EU GDPR | UK GDPR + DPA 2018
|
|---|---|---|
Scope for identifiable images | Personal data if a person can be identified directly or indirectly [1, Art. 4(1)] | Equivalent definition retained in UK law |
Household exemption | Outside scope for purely personal activities [1, Art. 2(2)(c)] | Equivalent exemption |
Freedom of expression exemptions | Member States must reconcile via Article 85 [1] | Implemented through DPA 2018 special purposes provisions [3] |
Video surveillance guidance | EDPB Guidelines 3/2019 on processing of personal data through video devices [2] | ICO guidance on video surveillance/CCTV and information rights [4] |
Children’s images | Children merit specific protection; apply a risk-based approach | ICO guidance highlights additional protections for children [5] |
For implementation questions or tailored deployment options, Contact us.
Relying on public-place capture is a misconception. The location of capture does not remove obligations if people are identifiable. Another frequent issue is masking only faces while leaving tattoos or name-printed clothing visible. Finally, do not rely solely on manual review at scale. Combining automated detection with a documented review step can produce more consistent outcomes.
Not always. If individuals are not identifiable due to distance or image quality, GDPR may not apply. This is context dependent and should be documented against Recital 26 criteria.
Logos are not personal data by themselves. If a logo is strongly linked to a person in a specific frame (for example, a named staff member in a small local business), consider whether other identifiers make the person identifiable. Otherwise, masking logos is usually a brand, confidentiality, or contractual issue rather than a data protection requirement.
Uniforms with names, numbers, or unique local markings can identify individuals. Masking those parts is a common compliance approach, while generic uniforms without identifiers often do not require masking on that basis alone.
Use conservative masking of faces and name tags unless an appropriate legal basis and safeguards are in place. Review images for school insignia and predictable re-identification through context.
Consent can be valid in some settings, but it must be freely given, specific, informed, and unambiguous, and it must be as easy to withdraw as to give. For broad public dissemination, consent can be hard to manage. Anonymization reduces identifiability and can lower compliance risk, but both approaches should align with purpose and context.
Yes. Detection accuracy is context dependent. Low light, angles, reflections, and occlusions can cause misses. A documented human review step is recommended for high-risk publications.
This article covers visual data anonymization. Where videos include audio, a separate assessment is needed for voices or spoken information that can identify someone.
