ETSI Domain: Data Management + Ethics

Crossover domain for ETSI metaverse categorisation addressing ical data handling, privacy-preserving storage, consent management, and responsible data governance.

Bridge-To

Semantic Classification

Content

Compositional Relationships (Components)

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:ConsentManagement))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:PrivacyControls))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:Anonymization))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:AuditLogging))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:PrivacyPreservingAnalytics))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:DataMinimisation))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:hasPart ai:RightToBeForgotten))

Dependency Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:requires ai:DataManagement))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:requires ai:EthicsAndLaw))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:requires ai:GDPR))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:requires ai:PrivacyByDesign))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:requires ai:DataProtectionImpactAssessment))

Capability Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:enables ai:PrivacyPreservingAnalytics))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:enables ai:UserControl))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:enables ai:ComplianceVerification))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:enables ai:DataSubjectRights))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:enables ai:TransparencyReporting))

Implementation Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:implements ai:ETSIMetaverseDomainTaxonomy))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:implements ai:GDPRCompliance))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:implements ai:EUAIActDataGovernance))

Reduction Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:reducesTo ai:AIGovernanceAndEthics))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:reducesTo ai:ETSIDomainDataManagement))

Support Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:supports ai:EUAIAct))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:supports ai:TrustworthyAI))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:supports ai:ResponsibleDataHandling))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:supports ai:DataGovernance))

Usage Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:uses ai:DifferentialPrivacy))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:uses ai:FederatedLearning))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:uses ai:HomomorphicEncryption))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:uses ai:Blockchain))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:uses ai:ZeroKnowledgeProofs))

Contrast Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:contrastsWith ai:ETSIDomain_DataMgmt_AI))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:contrastsWith ai:ETSIDomain_DataMgmt_Security))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:contrastsWith ai:ETSIDomain_DataMgmt_CreativeMedia))

Relation Relationships

SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:relatedTo ai:DataPrivacy))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:relatedTo ai:DigitalRights))
SubClassOf(ai:ETSIDomain_DataMgmt_Ethics
  ObjectSomeValuesFrom(ai:relatedTo ai:AIFairness))

About

The ETSI Domain: Data Management + Ethics crossover node addresses a challenge that has become increasingly acute as immersive and spatial computing platforms have matured: the data generated within metaverse environments is qualitatively different from, and far more sensitive than, most conventional digital data types. Spatial computing platforms capture gaze vectors, head orientation, hand tracking data, voice inputs, physiological signals from biosensors, and social proximity graphs — data modalities that can reveal emotional states, health conditions, cognitive processes, and social relationships that users may never have consciously consented to disclose. Research by Nair et al. (2023) demonstrated that motion data alone from VR headsets can uniquely identify individuals with high accuracy, raising fundamental questions about the adequacy of conventional anonymisation techniques for immersive environment data. ETSI’s crossover taxonomy node acknowledges this challenge by classifying metaverse components that must satisfy both data management functionality and ethical governance constraints as a distinct category requiring dedicated architectural attention.

The ethical data management framework for metaverse environments draws on a layered regulatory and normative architecture. At the European level, the GDPR (Regulation 2016/679) establishes foundational principles — lawfulness, fairness, transparency, purpose limitation, data minimisation, accuracy, storage limitation, integrity, confidentiality, and accountability — that apply fully to metaverse deployments. The EU AI Act (Regulation 2024/1689) adds specific obligations for AI systems that process personal data, including mandatory Data Protection Impact Assessments for high-risk AI systems, transparency obligations, and requirements for human oversight. At the UK level, the Data (Use and Access) Act 2025 updated the UK GDPR framework, introducing new lawful bases for data processing including data access for scientific research and public interest purposes, alongside the existing consent and legitimate interest bases. The UK Information Commissioner’s Office (ICO) updated its guidance on anonymisation in 2024–2025, providing case studies on differential privacy, pseudonymisation, and synthetic data — techniques that are core components of the Privacy-Preserving Analytics sub-domain covered by this ETSI classification.

The Consent Management component is particularly critical in metaverse contexts because the immersive, continuous nature of virtual environment participation means that consent must be dynamic, granular, and contextually sensitive rather than a one-time click-through agreement. ISO 29184 (Online Privacy Notices and Consent) provides a normative standard for consent management that ETSI-classified systems must align with. Blockchain technology has emerged as a technically promising mechanism for implementing immutable consent registries — ensuring that consent records cannot be retroactively modified and enabling decentralised verification of consent status across multi-platform metaverse ecosystems. ETSI’s bridging of this domain to the blockchain concept reflects this architectural pattern.

Components and Architecture

• Consent Management: The technical and governance infrastructure for capturing, storing, enforcing, and withdrawing user consent for specific data processing activities. In metaverse contexts, consent must be granular (per data type, per processing purpose, per recipient), contextually presented (within the immersive environment without breaking user experience), and dynamically enforceable (withdrawal of consent must propagate through all downstream processing pipelines). ISO 29184 provides the normative standard. Platforms include OneTrust, Usercentrics, and open-source CMP frameworks aligned with IAB Europe’s Transparency and Consent Framework (TCF) 2.2.

• Privacy Controls: Technical mechanisms that restrict data access and use to authorised purposes and authorised parties. In distributed metaverse architectures, privacy controls must enforce purpose limitation across a complex graph of platform components, third-party services, and edge computing nodes. Includes attribute-based access control (ABAC), data tagging and labelling for purpose limitation, and policy engines that evaluate access requests against consent records.

• Anonymization and Pseudonymisation: Data transformation techniques that reduce re-identification risk. The ICO’s updated anonymisation guidance (2024–2025) distinguishes: full anonymisation (irreversible, removes personal data entirely), pseudonymisation (reversible identifier replacement, still personal data under GDPR), k-anonymity, l-diversity, t-closeness, and Differential Privacy (adding calibrated noise to outputs). For metaverse data, traditional anonymisation techniques are insufficient for high-dimensional behavioural data; differential privacy and synthetic data generation are the leading privacy-enhancing technologies (PETs) for this context.

• Audit Logging: Tamper-evident recording of data access, processing operations, and consent decisions. Audit logs must be immutable (or at minimum integrity-protected), machine-readable, and retained for the regulatory period applicable to the processing activity. In AI systems, audit logging includes recording the data sources used for training, the model versions deployed, and the inference decisions made — directly connecting this component to the ETSI Domain: Data Management + AI domain’s Experiment Tracking and Model Registry components.

• Privacy-Preserving Analytics: Computational methods that enable statistical analysis and model training on sensitive data without exposing individual data subjects. The primary techniques are:

  • Differential Privacy: Adding calibrated Laplace or Gaussian noise to query outputs or gradient updates to bound the privacy loss for any individual. Adopted by Apple, Google, and the US Census Bureau for real-world deployments.
  • Federated Learning: Training machine learning models across distributed data sources without centralising raw data. The European Data Protection Supervisor published TechDispatch guidance on federated learning in June 2025, affirming its relevance for GDPR compliance.
  • Homomorphic Encryption: Computing on encrypted data without decryption, enabling privacy-preserving cloud analytics. Commercially relevant implementations include Microsoft SEAL and IBM HELib.
  • Secure Multi-Party Computation: Enabling multiple parties to jointly compute a function over their private inputs without revealing those inputs to each other.
  • Synthetic Data Generation: Creating statistically representative synthetic datasets that preserve analytical utility without containing real personal data.

• Data Minimisation: The architectural principle and technical enforcement mechanisms ensuring that only data necessary for the specified purpose is collected, processed, and retained. In metaverse contexts, data minimisation must be implemented at the sensor/capture layer (minimum required data types), the processing layer (purpose-limited transformation pipelines), and the storage layer (automatic retention enforcement).

• Right to be Forgotten (Right to Erasure): Technical mechanisms implementing GDPR Article 17 obligations to erase personal data on request. In immersive environments with complex data dependency graphs, right-to-erasure implementation requires capability to trace all derived data products (trained model contributions, cached analytics outputs, replicated records) that incorporate the subject’s data and either erase or re-derive them without that subject’s data.

  Ethical Data Governance Frameworks

  The ETSI Data Management + Ethics domain draws on several normative ethical governance frameworks:

• GDPR principles (Article 5): Lawfulness, fairness, transparency, purpose limitation, data minimisation, accuracy, storage limitation, integrity and confidentiality, accountability.

• UK Government Data and AI Ethics Framework (updated 2025): Provides a self-assessment tool for public sector organisations built on accountability, transparency, fairness, and the right to challenge automated decisions. Directly applicable to UK public sector metaverse deployments (NHS, education, civic services).

• ISO/IEC 38507:2022 (Governance of IT — Governance implications of the use of artificial intelligence by organisations): Maps ethical AI considerations to IT governance obligations.

• IEEE P7016 (Standard for Ethically Aligned Design and Operation of Metaverse Systems): Emerging standard specifically addressing metaverse ethical data handling obligations.

• ENISA Guidelines on Privacy-Enhancing Technologies: The EU Agency for Cybersecurity has published guidance on deploying differential privacy, pseudonymisation, and federated learning in compliance with GDPR — foundational references for the Privacy-Preserving Analytics component.

• ICO AI and Data Protection Guidance (2023–2025): The UK ICO’s updated guidance covers fairness, transparency, and lawful basis for AI-driven data processing; its risk toolkit provides structured assessment of AI systems’ compliance with UK GDPR.

  Use Cases in Metaverse Contexts

• Avatar identity and biometric data governance: Metaverse platforms that capture facial expressions, body gestures, and voice for avatar animation process special category data under GDPR Article 9. The Data Management + Ethics domain provides the classification for systems implementing consent gating, purpose limitation, and right-to-erasure for such biometric data streams.

• Spatial behaviour analytics with differential privacy: Platforms that wish to analyse aggregate user movement patterns within virtual environments without exposing individual trajectories must implement Differential Privacy on their analytics pipelines. This use case sits squarely within the Privacy-Preserving Analytics component of this domain.

• Federated AI training on distributed user data: A metaverse platform that trains personalisation models on data distributed across user devices without centralising raw data implements Federated Learning — a key Privacy-Preserving Analytics technique. The European Data Protection Supervisor’s 2025 TechDispatch on federated learning provides GDPR-aligned guidance for this pattern.

• Cross-platform consent management with blockchain audit trail: As metaverse ecosystems span multiple platforms and jurisdictions, Blockchain-based consent registries enable immutable, decentralised consent tracking with cryptographic proof of consent status — implementing the Consent Management and Audit Logging components simultaneously.

• Right-to-erasure in AI-powered virtual environments: When a user requests deletion of their data from a metaverse platform that has used their interaction data to train personalisation models, the platform must trace the data subject’s contribution through training pipelines and either retrain models without that data (machine unlearning) or demonstrate that the trained model provides sufficient differential privacy guarantees to satisfy the erasure obligation.

• Children’s data protection in educational XR: Extended reality educational platforms that deploy immersive learning environments for children under 16 face heightened consent requirements (parental consent, Article 8 GDPR) and must implement correspondingly strict data minimisation and purpose limitation controls — a canonical implementation of the Data Minimisation and Consent Management components.

  Data Subject Rights Implementation

  A core operational requirement of the ETSI Data Management + Ethics domain is the technical implementation of data subject rights as defined under the GDPR (Articles 15–22) and UK GDPR. In metaverse environments, these rights present distinctive implementation challenges due to the complexity and scale of data dependencies:

  Right to Access (Article 15): Data subjects may request confirmation of whether their personal data is being processed, access to the data itself, and information about processing purposes, retention periods, and recipients. In a metaverse platform capturing hundreds of data streams per user session, implementing this right requires a comprehensive personal data inventory linked to a data subject identifier graph — a capability provided by the Consent Management and Audit Logging components.

  Right to Erasure / Right to be Forgotten (Article 17): Data subjects may request erasure of personal data in specified circumstances. In AI-powered metaverse platforms that have used a user’s data for model training, the right to erasure creates a machine unlearning problem: the trained model may embody the erased user’s data contribution in its weights. Three technical approaches exist: (a) full model retraining excluding the erased user’s data (computationally prohibitive at scale); (b) machine unlearning algorithms that surgically remove the user’s data contribution without full retraining (active research area at Cambridge, Edinburgh, and UCL); (c) relying on differential privacy guarantees to argue that the model provides sufficient statistical anonymity to satisfy the erasure obligation (legally contested approach). The ETSI domain’s Right to be Forgotten component must implement one or more of these approaches.

  Right to Portability (Article 20): Data subjects may request their personal data in a machine-readable format suitable for transfer to another controller. In metaverse contexts, portability of user data — including avatar configurations, social graph data, spatial preference data, and behavioural history — aligns with emerging metaverse interoperability standards from the Open Metaverse Foundation and the Metaverse Standards Forum, which seek to enable user data portability across platform boundaries.

  Right to Restrict Processing (Article 18): Data subjects may request restriction of processing in certain circumstances, such as when data accuracy is contested. Implementing processing restriction in a complex AI-powered metaverse platform requires the capability to flag individual user data records and exclude them from downstream processing pipelines — a capability integrated into the Data Pipelines and Feature Store components.

  Right to Object to Automated Decision-Making (Article 22): Data subjects may object to decisions based solely on automated processing that significantly affect them. In AI-powered metaverse platforms, this right applies to algorithmic content moderation, access control decisions, and personalisation systems. Implementing this right requires human oversight mechanisms and the ability to invoke human review for contested automated decisions.

  Privacy-Enhancing Technologies (PETs) Technical Reference

  The Privacy-Preserving Analytics component of this domain encompasses a taxonomy of Privacy-Enhancing Technologies (PETs) with distinct technical properties and applicability profiles:

  Differential Privacy (DP): Provides a mathematical guarantee (epsilon, delta)-differential privacy bounding the information leakage about any individual in a dataset. Two deployment modes: (a) Local DP — each data point is perturbed before leaving the source device (used by Apple for telemetry, Google for Chrome DP reports); (b) Central DP — the data curator adds noise to query outputs or gradient updates after centralising data (stronger utility at the same privacy budget). In metaverse contexts, DP is applied to: aggregate spatial analytics (movement heatmaps), ML model gradient updates during training, and statistical summaries of biometric sensor data. Key parameter: the privacy budget epsilon (lower = stronger privacy, higher = better utility). The ICO’s 2025 case studies include a UK bank applying mixed-noise differential privacy to transaction data for market analysis.

  Federated Learning (FL): Trains ML models across decentralised data sources without centralising raw data. The coordinator sends the current model to participating clients; each client computes gradient updates on its local data; clients send gradient updates (not raw data) back to the coordinator; the coordinator aggregates updates (e.g., FedAvg algorithm). Residual risks include gradient inversion attacks (reconstructing training data from gradient updates) and membership inference attacks — addressed by combining FL with DP (DP-FL) or secure aggregation. The EDPS TechDispatch of June 2025 affirms FL’s GDPR relevance while noting these residual risks.

  Homomorphic Encryption (HE): Enables computation on ciphertext, producing encrypted outputs that match the result of computing on plaintext. Supports GDPR compliance for cloud analytics by keeping data encrypted throughout processing. Current performance overhead: approximately 10–100x vs. plaintext computation (Microsoft SEAL, IBM HELib implementations). Hardware-accelerated HE (Intel HERACLES, FAB consortium accelerators) is targeting 3–5x overhead, making real-time HE analytics feasible by 2027–2028.

  Secure Multi-Party Computation (SMPC): Multiple parties jointly compute a function over their private inputs without revealing inputs to each other, using techniques including garbled circuits, secret sharing, and oblivious RAM. Applicable to metaverse scenarios involving multiple platform operators computing joint analytics on their combined user bases without sharing raw user data.

  Synthetic Data Generation: Creating statistically representative artificial datasets using generative models (VAEs, GANs, Diffusion Models) that replicate the statistical properties of real datasets without containing actual personal data. ICO guidance confirms that properly generated synthetic data with appropriate evaluation can be treated as anonymous data outside the scope of GDPR. Leading UK tools include Gretel.ai, Mostly AI, and Statice (now Anonos).

  K-Anonymity, L-Diversity, T-Closeness: Classical statistical disclosure control techniques ensuring that each data record is indistinguishable from at least k-1 other records (k-anonymity), or that sensitive attributes have sufficient diversity within equivalence classes (l-diversity), or that the distribution of sensitive attributes within classes matches the overall distribution (t-closeness). ICO guidance covers these techniques alongside DP and synthetic data.

  Zero-Knowledge Proofs (ZKPs): Cryptographic protocols enabling one party to prove to another that a statement is true without revealing any information beyond the truth of the statement. Applicable to consent verification (proving consent was granted without revealing consent record contents) and identity verification (proving age or eligibility without revealing identity). Blockchain-integrated ZKP systems are being deployed in decentralised identity frameworks relevant to metaverse consent management.

  Relationship to ETSI Domain: Data Management + AI

  The ETSI Data Management + Ethics and ETSI Domain: Data Management + AI domains are architecturally complementary and operationally coupled. Their relationship is best understood as two orthogonal governance layers applied to the same underlying Data Management infrastructure:

  The Data Management + AI domain governs the operational efficiency, reproducibility, and deployment lifecycle of AI/ML systems — concerned with training data quality as an accuracy and performance concern, model versioning as a deployment reliability concern, and experiment tracking as a reproducibility concern.

  The Data Management + Ethics domain governs the ethical legitimacy, legal compliance, and rights-respecting character of the same data management activities — concerned with training data quality as a fairness and bias concern, data retention as a minimisation and proportionality concern, and processing records as a consent accountability concern.

  Concretely, the Experiment Tracking component of the AI domain and the Audit Logging component of the Ethics domain both maintain records of data processing events — but the AI domain’s focus is on reproducibility and performance, while the Ethics domain’s focus is on consent accountability and data subject rights. Similarly, the Training Data Repository of the AI domain enforces version coherence for model reproducibility, while the Ethics domain requires the same repository to enforce data minimisation, consent scope, and right-to-erasure obligations. Organisations deploying metaverse AI systems must implement both governance layers simultaneously, using the ETSI crossover taxonomy to identify which components are required by each regulatory and technical dimension.

  Metaverse-Specific Privacy Risks and Mitigations

  Metaverse and extended reality environments introduce novel privacy risks not fully addressed by existing GDPR guidance, driving the need for the ETSI Data Management + Ethics crossover domain:

  Behavioural biometric identification: Research by Nair et al. (2023) demonstrated that VR motion data — head and controller movements — can uniquely identify individuals with accuracy exceeding 95% after less than 100 seconds of interaction. This makes conventional anonymisation of motion data insufficient; differential privacy or data minimisation at the capture layer are required.

  Gaze-based sensitive attribute inference: Eye-tracking data captured for foveated rendering can reveal cognitive load, emotional state, sexual orientation, and health conditions. GDPR Article 9 special category data provisions apply. The Data Management + Ethics domain’s Data Minimisation component must ensure that gaze data is processed only for the specified rendering purpose and not retained or repurposed.

  Social graph construction from spatial proximity: Metaverse platforms that track which avatars are co-present in shared virtual spaces can construct detailed social graphs, revealing relationships, associations, and behavioural patterns. The Privacy Controls component must implement purpose limitation and consent scope enforcement to prevent secondary use of spatial co-presence data for profiling.

  Cross-platform data integration and identity linkage: Users who interact with multiple metaverse platforms using consistent avatar identifiers or biometric signatures may be re-identified across platforms by data brokers or platform operators with access to multiple datasets. The Anonymization component must assess cross-platform re-identification risk when processing data from multi-platform metaverse ecosystems.

  Children in virtual environments: The prevalence of minors in gaming and educational metaverse platforms triggers heightened data protection obligations (GDPR Article 8, UK Children’s Code). The Consent Management component must implement age verification, parental consent workflows, and stricter data minimisation defaults for identified minor users.

  Academic Context

  The academic foundations of the ETSI Data Management + Ethics domain draw from interdisciplinary research spanning computer science, law, philosophy, and social science:

• Privacy-by-design: Cavoukian (2009) formalised the seven privacy-by-design principles that underpin the architectural approach advocated by this domain — privacy as default, full functionality, end-to-end security, visibility and transparency, respect for user privacy. These principles were subsequently incorporated into GDPR Article 25 (data protection by design and by default) and the ICO’s accountability guidance, making them legally operative in EU and UK jurisdictions.

• Differential privacy theory: Dwork et al. (2006) established the mathematical foundations of differential privacy, providing the formal guarantees that underpin the Differential Privacy component’s privacy claims. Subsequent work by Abadi et al. (2016) adapted DP to deep learning via the DP-SGD algorithm, enabling practical differential privacy for neural network training.

• Federated learning: McMahan et al. (2017) introduced federated learning at Google, demonstrating privacy-preserving collaborative training at scale. Subsequent work on combining FL with differential privacy (DP-FL) and secure aggregation addressed the residual gradient leakage risks in the original FL protocol.

• Metaverse privacy risks: Nair et al. (2023) demonstrated unique identification from VR motion data, motivating the ETSI domain’s emphasis on advanced privacy-enhancing technologies beyond conventional anonymisation. Earlier work by Narayanan and Shmatikov (2008) on de-anonymisation of the Netflix Prize dataset established the inadequacy of naive anonymisation for high-dimensional behavioural data.

• Ethical AI frameworks: Mittelstadt et al. (2016) surveyed the ethics of algorithms; Floridi et al. (2018) published the AI4People ethical framework adopted by the High-Level Expert Group on AI; Jobin et al. (2019) provided a global convergence analysis of AI ethics principles across 84 documents, identifying transparency, justice/fairness, non-maleficence, responsibility, and privacy as the most commonly cited principles — all directly instantiated in the components of this ETSI domain.

• Legal basis and consent research: Wachter et al. (2017) analysed the right to explanation under GDPR; Goodman and Flaxman (2017) examined algorithmic decision-making and GDPR obligations; Shokri et al. (2017) demonstrated membership inference attacks against ML models, empirically motivating the need for differential privacy in training data management.

  Current Landscape (2026)

  The regulatory and technical landscape in 2026 has significantly elevated the operational importance of this domain:

• EU AI Act enforcement milestones: From February 2025, prohibitions on unacceptable-risk AI (including certain biometric categorisation) came into force. From August 2025, GPAI model obligations became binding. From August 2026, the full Article 10 data governance obligations for high-risk AI apply — making the data management ethics components (training data quality, representativeness, bias assessment) legally mandatory.

• GDPR enforcement maturation: EU data protection authorities have issued significant fines (including meta-platform fines exceeding €1 billion) for inadequate consent management and data minimisation practices in cross-platform digital services. These enforcement actions directly shape how metaverse platforms must implement this domain’s components.

• Data (Use and Access) Act 2025 (UK): The UK’s updated data protection legislation, passed in 2025, introduced new lawful bases and streamlined the consent framework while strengthening ICO enforcement powers. The ICO’s updated anonymisation guidance (2024–2025) specifically addresses differential privacy and synthetic data as GDPR-compatible privacy-enhancing technologies.

• EDPS TechDispatch on Federated Learning (June 2025): The European Data Protection Supervisor published dedicated guidance on federated learning, affirming its potential for GDPR compliance while identifying residual risks including gradient inversion attacks and membership inference — considerations that must be addressed in system designs implementing the Privacy-Preserving Analytics component.

• ETSI EN 304 223 (AI security baseline): ETSI’s global baseline standard for AI security, published 2025–2026, integrates with the Data Management + Ethics domain by establishing security requirements for personal data processed in AI systems — including encryption, access control, and breach notification obligations that implement the Privacy Controls and Audit Logging components.

• Blockchain-based consent registries: Production deployments of blockchain-backed consent management have emerged in European digital health (HL7 FHIR-compatible consent tokens on permissioned blockchains) and legal tech contexts, demonstrating the technical viability of the Blockchain bridge concept in this domain.

  UK Context

  The United Kingdom has a distinctive regulatory and institutional landscape that shapes how this domain is implemented in practice:

• Information Commissioner’s Office (ICO): As the UK’s lead data protection authority and de facto primary AI regulator (given that data protection law touches virtually every AI system), the ICO’s updated guidance on anonymisation (2024–2025), AI and data protection risk toolkit, and forthcoming statutory code of practice on AI and automated decision-making directly instantiate the normative requirements of this domain. The ICO’s approach to differential privacy and pseudonymisation — as evidenced in published case studies on a bank applying mixed-noise differential privacy for market analysis — is the authoritative UK reference for implementing Privacy-Preserving Analytics.

• UK Government Data and AI Ethics Framework (updated December 2025): The Responsible Data and AI team in the Cabinet Office launched an updated Data and AI Ethics Framework and self-assessment tool enabling public sector teams to innovate responsibly. This framework explicitly covers consent management, data minimisation, accountability, and transparency — the core components of the ETSI Data Management + Ethics domain — and is mandatory guidance for UK central government AI deployments.

• Centre for Data Ethics and Innovation (CDEI): The CDEI provides cross-government guidance on responsible data and AI innovation, publishing case studies and practical tools that operationalise the ethical principles of this ETSI domain in UK public sector contexts.

• NHS AI governance: NHS England’s ambition to be “the world’s most AI-enabled health service” is accompanied by AI governance policies requiring Data Protection Impact Assessments, consent management, and audit trails for all AI systems handling patient data — direct implementations of this domain’s components at national health system scale. Clinical AI deployments at NHS trusts in Leeds, Manchester, Sheffield, and Newcastle are subject to these governance requirements.

• BCS (The Chartered Institute for IT): BCS has highlighted growing demand for transparency, accountability, and ethical leadership in the tech sector, and its 2025 report on AI ethics and professional registrations in the UK identifies data ethics as a critical skill gap. BCS professional standards align with this ETSI domain’s emphasis on accountable, transparent, and consent-respecting data management.

• Alan Turing Institute: The UK’s national institute for data science and AI has published research on privacy-enhancing technologies, responsible data sharing, and federated learning frameworks directly applicable to the technical components of this domain. The institute’s work on differential privacy for public sector data sharing (e.g., ONS statistical disclosure control) provides concrete UK implementation evidence.

• Northern England industrial context: The NHS digitisation programmes in Manchester (Greater Manchester Health and Care Partnership), Leeds (West Yorkshire and Harness), and Newcastle (Newcastle Hospitals NHS Foundation Trust) represent significant deployments of health AI requiring all components of this domain — consent management for patient data, privacy-preserving analytics for population health insights, audit logging for clinical AI decisions, and data minimisation for sensitive health records.

  Future Directions (2026–2030)

• Machine unlearning at scale: As right-to-erasure obligations are applied to AI systems trained on personal data, machine unlearning — the ability to remove a data subject’s contribution from a trained model without full retraining — will become a critical technical requirement. Research at Cambridge, Edinburgh, and UCL is advancing machine unlearning algorithms including approximate unlearning via influence functions, exact unlearning via SISA training (sharded, isolated, sliced, and aggregated), and certified unlearning via differential privacy composition — providing a technical toolkit for implementing the Right to be Forgotten component in AI-powered metaverse platforms at scale.

• Homomorphic encryption for metaverse analytics: As HE performance improves (current Microsoft SEAL benchmarks show 10–100x overhead vs. plaintext; FAB consortium accelerators and Intel HERACLES target 3–5x overhead by 2028), practical deployment of HE-backed analytics in metaverse environments will become feasible, enabling genuinely privacy-preserving real-time analytics on biometric data streams without decryption at any processing stage.

• Decentralised consent management with verifiable credentials: The convergence of Blockchain, Decentralised Identity (W3C DID Core 1.0 standard, ISO/IEC 18013-5 mDL), and Verifiable Credentials (W3C VC Data Model 2.0) with Consent Management will enable users to hold portable, self-sovereign consent records across metaverse platforms — eliminating per-platform re-consent overhead and enabling cryptographically verified consent status checks without centralised consent registry dependency.

• Synthetic data maturation: Advances in generative AI — VAEs, diffusion models, consistency models, neural radiance fields for 3D spatial data synthesis — will make high-fidelity synthetic metaverse behavioural datasets routinely usable for model training without privacy risk, fundamentally changing the economics of the Privacy-Preserving Analytics component and enabling model training on data types (biometric, health, children’s behavioural data) that are otherwise too sensitive to use in centralised training pipelines.

• Real-time differential privacy for streaming data: Current differential privacy implementations impose offline or batch-mode constraints (offline query release, batch gradient computation in DP-FL). Research into streaming differential privacy — applying calibrated noise to real-time event streams via sliding window mechanisms — will enable privacy-preserving live analytics of metaverse avatar interactions, spatial heatmaps, and social graph dynamics without batch processing latency.

• EU AI Act Article 10 tooling ecosystem: The August 2026 enforcement deadline for Article 10 data governance obligations is driving a commercial tooling ecosystem for training data quality assessment, bias detection, representativeness testing, and dataset documentation (data cards, datasheets) — capabilities being integrated into Training Data Repository and Experiment Tracking platforms by MLflow, lakeFS, and cloud-native ML platforms.

• Metaverse-specific consent UX patterns: Research into consent interfaces that are non-disruptive within immersive 3D environments is generating emerging standards — spatial consent dialogues delivered as diegetic UI elements within virtual spaces, avatar-mediated consent confirmations using natural gesture interactions, gaze-based consent acknowledgement with eye-tracking confirmation, and temporal consent reminders integrated into virtual world transitions — that will mature into standardised UX patterns enabling the Consent Management component to be implemented without degrading immersive experience quality.

• Privacy-preserving federated analytics: Beyond federated learning for model training, federated analytics — computing aggregate statistics across distributed data without centralising individual records — is emerging as a privacy-preserving alternative to centralised analytics pipelines. Google’s Private Join and Compute, Apple’s Private Federated Statistics, and the OpenDP library represent the current toolkit for this trajectory, relevant to metaverse platforms seeking population-level insights without individual data centralisation.

• Algorithmic impact assessments: Building on DPIA frameworks, AI-specific algorithmic impact assessments (AIAs) examining fairness, bias, transparency, and accountability of deployed AI systems will become standard governance practice — extending the Audit Logging and Compliance Verification components to support structured ethical risk evaluation across the AI system lifecycle.

  Research and Literature

  1. Cavoukian, A. (2009). Privacy by design: The 7 foundational principles. Information and Privacy Commissioner of Ontario, Canada.
  2. Dwork, C., McSherry, F., Nissim, K., & Smith, A. (2006). Calibrating noise to sensitivity in private data analysis. Proceedings of the Theory of Cryptography Conference (TCC), Springer, LNCS 3876, 265–284.
  3. McMahan, H. B., Moore, E., Ramage, D., Hampson, S., & Agüera y Arcas, B. (2017). Communication-efficient learning of deep networks from decentralized data. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS).
  4. Nair, V., Gauthier, J. F., Sherrill, J. S., & Bailenson, J. N. (2023). Going incognito in the metaverse: Identifying and protecting against avatar fingerprinting. Proceedings of the IEEE VR 2023 Conference.
  5. Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016). The ethics of algorithms: Mapping the debate. Big Data & Society, 3(2), 2053951716679679.
  6. Floridi, L., Cowls, J., Beltrametti, M., Chatila, R., Chazerand, P., Dignum, V., Luetge, C., Madelin, R., Pagallo, U., Rossi, F., Schafer, B., Valcke, P., & Vayena, E. (2018). AI4People — An ethical framework for a good AI society. Minds and Machines, 28(4), 689–707.
  7. Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389–399.
  8. Gebru, T., Morgenstern, J., Vecchione, B., Wortman Vaughan, J., Wallach, H., Daumé III, H., & Crawford, K. (2018). Datasheets for datasets. Communications of the ACM, 64(12), 86–92.
  9. Bonawitz, K., Ivanov, V., Kreuter, B., Marcedone, A., McMahan, H. B., Patel, S., Ramage, D., Segal, A., & Seth, K. (2017). Practical secure aggregation for privacy-preserving machine learning. Proceedings of the 2017 ACM SIGSAC Conference on CCS, 1175–1191.
  10. European Parliament and Council. (2016). Regulation (EU) 2016/679 — General Data Protection Regulation. Official Journal of the European Union.
  11. European Parliament and Council. (2024). Regulation (EU) 2024/1689 on Artificial Intelligence (EU AI Act). Official Journal of the European Union.
  12. UK Parliament. (2025). Data (Use and Access) Act 2025. His Majesty’s Stationery Office, London.
  13. ICO. (2024–2025). Anonymisation, pseudonymisation, and privacy-enhancing technologies: Guidance and case studies. Information Commissioner’s Office, UK.
  14. EDPS. (2025, June 10). TechDispatch #1/2025 — Federated Learning. European Data Protection Supervisor.
  15. ENISA. (2022). Data protection engineering: From theory to practice. European Union Agency for Cybersecurity.
  16. ISO 29184:2020. (2020). Online privacy notices and consent. International Organization for Standardization.
  17. ISO/IEC 38507:2022. (2022). Information technology — Governance of IT — Governance implications of the use of artificial intelligence by organisations. ISO/IEC.
  18. Wachter, S., Mittelstadt, B., & Russell, C. (2017). Counterfactual explanations without opening the black box: Automated decisions and the GDPR. Harvard Journal of Law & Technology, 31(2), 841–887.
  19. Shokri, R., Stronati, M., Song, C., & Shmatikov, V. (2017). Membership inference attacks against machine learning models. Proceedings of the IEEE Symposium on Security and Privacy (S&P), 3–18.
  20. UK Government. (2025, December). Data and AI Ethics Framework and self-assessment tool. Cabinet Office / Central Digital and Data Office.
  21. BCS. (2025). AI ethics and professional registrations in the UK. The Chartered Institute for IT, London.
  22. Goodman, B., & Flaxman, S. (2017). European Union regulations on algorithmic decision-making and a “right to explanation.” AI Magazine, 38(3), 50–57.
  23. ETSI EN 304 223 V1.1.1 (2026). Securing Artificial Intelligence (SAI): Global baseline for AI security requirements. European Telecommunications Standards Institute.
  24. Abadi, M., Chu, A., Goodfellow, I., McMahan, H. B., Mironov, I., Talwar, K., & Zhang, L. (2016). Deep learning with differential privacy. Proceedings of the 2016 ACM SIGSAC Conference on CCS, 308–318.
  25. Ryffel, T., Trask, A., Dahl, M., Wagner, B., Mancuso, J., Rueckert, D., & Passerat-Palmbach, J. (2018). A generic framework for privacy preserving deep learning. arXiv:1811.04017.
  26. Kaissis, G. A., Makowski, M. R., Rückert, D., & Braren, R. F. (2020). Secure, privacy-preserving and federated machine learning in medical imaging. Nature Machine Intelligence, 2(6), 305–311.
  27. Narayanan, A., & Shmatikov, V. (2008). Robust de-anonymization of large sparse datasets. Proceedings of the IEEE Symposium on Security and Privacy (S&P), 111–125.
  28. ICO. (2025). AI and data protection risk toolkit. Information Commissioner’s Office, UK. https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/

  Key Terminology

• Differential Privacy (DP): A mathematical framework for bounding the information leakage about any individual from the output of a data analysis or ML training process. Characterised by privacy parameters epsilon (privacy loss bound) and delta (failure probability). Lower epsilon = stronger privacy; higher epsilon = better utility.

• Federated Learning (FL): A distributed ML training paradigm in which model training occurs across decentralised data sources (client devices, institutional servers) without centralising raw data. The coordinator aggregates gradient updates, not raw data.

• Homomorphic Encryption (HE): A cryptographic scheme enabling arithmetic operations on ciphertext that produce results matching those on plaintext after decryption. Enables cloud computation on encrypted personal data without decryption.

• Consent Management: The technical and governance infrastructure for capturing, recording, enforcing, and withdrawing user consent for specific data processing purposes, in compliance with GDPR Article 7 and UK GDPR equivalents.

• Data Minimisation: The GDPR principle (Article 5(1)(c)) requiring that personal data be adequate, relevant, and limited to what is necessary for the specified processing purpose. Implemented technically through purpose-scoped data collection, processing limits, and automatic retention enforcement.

• Anonymisation: The irreversible process of transforming personal data such that the data subject can no longer be identified, directly or indirectly, placing the resulting data outside the scope of GDPR. Contrasted with pseudonymisation, which is reversible.

• Pseudonymisation: Replacement of direct identifiers with pseudonyms in a way that the original data subject can be re-identified using separately stored information (the key). Personal data under GDPR; provides limited but useful risk reduction for storage and processing.

• Privacy by Design: An architectural approach incorporating privacy protections as foundational design requirements from the earliest stage of system design, rather than as afterthought additions. Codified in GDPR Article 25.

• Data Protection Impact Assessment (DPIA): A formal risk assessment required under GDPR Article 35 for processing activities likely to result in high risks to individuals’ rights and freedoms. Mandatory for high-risk AI systems under the EU AI Act.

• Privacy-Preserving Analytics: Analytics methods that derive statistical insights from sensitive data without exposing individual data subjects, including differential privacy, federated learning, secure multi-party computation, and synthetic data generation.

• Machine Unlearning: Algorithms that remove the influence of specified training data from a trained model without full retraining, implementing the Right to Erasure for AI systems. Active research area at Cambridge, Edinburgh, and UCL.

• K-Anonymity: A property of a dataset where each record is indistinguishable from at least k-1 other records with respect to specified quasi-identifier attributes. Provides protection against identity disclosure.

• Right to Erasure: GDPR Article 17 right enabling data subjects to request deletion of their personal data in specified circumstances. In AI systems, creates the machine unlearning problem — how to remove a data subject’s contribution from a trained model.

• Audit Logging: Tamper-evident, machine-readable records of data processing operations, consent decisions, and access events, providing accountability evidence for regulatory compliance and data subject rights enforcement.

• Blockchain Consent Registry: An immutable ledger implementation for storing consent records, enabling decentralised verification of consent status across multi-platform metaverse ecosystems with cryptographic proof of consent lineage.

• Zero-Knowledge Proof (ZKP): A cryptographic protocol enabling one party to prove a statement’s truth to another without revealing information beyond the statement’s validity. Used in consent verification and identity attestation without personal data disclosure.

• Synthetic Data: Artificially generated data that preserves the statistical properties of a real dataset without containing actual personal data. ICO guidance (2024–2025) affirms that properly generated synthetic data can qualify as anonymous data outside GDPR scope.

• Transparency Reporting: The practice of publishing information about data processing activities, model performance, and audit outcomes to enable public accountability and regulatory oversight of AI systems.

• Privacy Budget: In differential privacy, the accumulated privacy loss (epsilon) consumed by a sequence of queries or training steps on a dataset. Managing privacy budgets prevents over-querying depleting privacy protections.

• Granular Consent: Consent captured at fine-grained levels of specificity — per data type, per processing purpose, per data recipient, per retention period — enabling data subjects to selectively permit specific processing activities while refusing others.

  Cross-Domain Integration Patterns

  The ETSI Data Management + Ethics domain integrates with other ETSI metaverse taxonomy nodes through several architectural patterns:

  Integration with ETSI Domain: Data Management + AI: AI operational concerns (training data quality, model reproducibility, experiment tracking) must be implemented simultaneously with ethical governance concerns (consent scope, data minimisation, right to erasure). The Training Data Repository must enforce both technical version coherence (AI concern) and consent-scoped access controls (ethics concern). The Experiment Tracking component’s audit records serve double duty as technical reproducibility records (AI concern) and processing accountability records (ethics concern).

  Integration with ETSI Domain: Data Management + Security: Privacy and security are complementary but distinct concerns. Security controls (encryption at rest and in transit, access control, intrusion detection) are necessary but not sufficient for privacy compliance. The ethics domain’s Privacy Controls operate at a higher semantic level — enforcing purpose limitation and consent scope — while the security domain’s controls operate at the infrastructure level. Both are required for a compliant metaverse data management implementation.

  Integration with ETSI Domain Security and Privacy: ETSI’s Security and Privacy domain provides the cryptographic and security infrastructure (including Blockchain, Homomorphic Encryption, Zero-Knowledge Proofs) that enables the Privacy-Preserving Analytics and Blockchain consent registry components of the Data Management + Ethics domain.

  Integration with ETSI Domain: Data Management: The base data management domain’s storage and processing infrastructure must be configured with appropriate privacy controls (encryption, access logging, retention enforcement) before the ethics-specific components can be layered on top. The Data Lifecycle management component of the base domain is directly extended by the Data Management + Ethics domain’s data minimisation and storage limitation enforcement.

  Integration with ETSI Domain AI + Governance: The AI governance domain provides the broader ethical AI framework within which the Data Management + Ethics crossover domain’s components are situated — covering algorithmic fairness, transparency, explainability, and accountability obligations that apply at the AI system level rather than specifically at the data management infrastructure level.

  Freshness Updates (2024–2026)

• December 2025: UK Government launched updated Data and AI Ethics Framework and self-assessment tool (Cabinet Office / Central Digital and Data Office), providing structured self-assessment for public sector organisations implementing data ethics governance aligned with this ETSI domain.

• June 2025: EDPS published TechDispatch #1/2025 on Federated Learning, affirming FL’s GDPR compliance relevance while noting residual risks (gradient inversion, membership inference) requiring DP-FL or secure aggregation mitigations.

• 2025: UK Data (Use and Access) Act 2025 passed, updating the UK GDPR framework with new lawful bases for data processing, strengthened ICO enforcement powers, and updated provisions for scientific research data access.

• 2024–2025: ICO published updated anonymisation and pseudonymisation guidance with case studies covering differential privacy, synthetic data, and homomorphic encryption as GDPR-compatible privacy-enhancing technologies.

• February 2025: EU AI Act prohibition provisions (including on certain biometric categorisation systems) came into force, directly affecting metaverse platforms deploying AI-powered avatar categorisation or emotion recognition.

• August 2025: EU AI Act GPAI model obligations binding; training data transparency summaries required for general-purpose AI providers.

• August 2026: EU AI Act Article 10 data governance obligations fully binding on high-risk AI providers; Data Protection Impact Assessment required for AI systems processing biometric, health, or children’s data.

• 2026: ETSI EN 304 223 (global baseline for AI security) published, integrating data protection obligations with AI security requirements in a unified technical specification relevant to this domain’s Privacy Controls and Audit Logging components.

• 2025–2026: European AI Liability Directive negotiations advanced, proposing strict liability for damages caused by AI systems in high-risk categories and requiring traceable Audit Logging and Transparency Reporting from deployed AI operators — directly enforcing the accountability components of this ETSI domain through civil liability mechanisms.

• 2026: BCS published updated AI ethics and professional registrations report identifying data ethics competence (covering Consent Management, Data Minimisation, and Privacy Controls) as the highest-priority professional skill gap in the UK computing sector.

• 2026: Metaverse Standards Forum working groups on ethics and user rights have advanced guidelines for consent UX patterns and data portability in immersive environments, directly informing the next generation of Consent Management tooling and interaction design standards aligned with this ETSI domain.

Provenance