Narrative Goldmine

ETSI Domain AI

52 min read

Properties

Type
  • Class
Resource
  • urn:ngm:class:etsi-domain-ai
Domain
  • artificial-intelligence
Status
  • stable
Maturity
  • established
Quality
  • 0.35
Is a
Part of
Has part
Requires
Enables
Generated
  • by: process:vault-migrate/1.0 · at: 2026-09-22T12:41:39.024503178Z
Bridges to
Contrasts with
Depends on
Implemented in layer
Implements
Legacy term id
  • 20330
Links
Page resource
  • urn:visionflow:page:babbbc5e48be6ec89517066976a3c2d1e23b129b76c632478d3b7928a3f9ad6d
Related to
Standardized by
Supports
Uses

Domain marker concept for categorising metaverse components related to artificial intelligence, machine learning, and computational intelligence capabilities.

Bridge-To

Semantic Classification

Content

Compositional Relationships (Components)

SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:hasPart ai:ETSIDomainAICreativeMedia))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:hasPart ai:ETSIDomainAIDataMgmt))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:hasPart ai:ETSIDomainAIGovernance))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:hasPart ai:ETSIDomainAIHumanInterface))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:hasPart ai:AIServiceClassification))

Dependency Relationships

SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:requires ai:TechnicalCommittee))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:requires ai:ETSIISGENIGroup))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:requires ai:ETSITCSAIGroup))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:dependsOn ai:ComputationAndIntelligenceDomain))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:dependsOn ai:ETSIDomainTaxonomy))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:dependsOn ai:MachineLearning))

Capability Relationships

SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:enables ai:AIServiceClassification))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:enables ai:IntelligenceLayerCategorization))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:enables ai:StandardsGapAnalysis))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:enables ai:AIConformityAssessment))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:supports ai:EUAIAct))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:supports ai:HarmonisedStandards))

Implementation Relationships

SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:implements ai:AIStandardisation))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:implements ai:TaxonomyGovernance))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:uses ai:MachineLearning))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:uses ai:DeepLearning))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:uses ai:NaturalLanguageProcessing))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:uses ai:ComputerVision))

Reduction Relationships

SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:reducesTo ai:AIResearchArea))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:reducesTo ai:ComputationAndIntelligenceDomain))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:reducesTo ai:DomainClassificationConcept))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:reducesTo ai:StandardisationOntologyNode))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:bridges ai:EUAIActDomainClassification))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:bridges ai:ISOIECJtc1Sc42AIVocabulary))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:relatedTo ai:ITUTAIStandards))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:relatedTo ai:NLPDomain))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:contrastsWith ai:NISTAIRiskManagementFramework))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:contrastsWith ai:ISOIEC42001ManagementSystem))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:standardizedBy ai:ETSI))
SubClassOf(ai:ETSIDomainAI
  ObjectSomeValuesFrom(ai:implements ai:ETSIWorkProgramme))

About

ETSI Domain AI is the canonical root domain concept by which the European Telecommunications Standards Institute organises, classifies, and cross-references all work items, technical specifications, group reports, and standards that have artificial intelligence as a defining characteristic. The concept does not itself constitute a standard but instead forms the structural anchor in the ETSI Domain Taxonomy from which a family of cross-domain intersection markers branch outward into the application landscape: ETSI Domain AI + Creative Media covers AI applied to generative content, procedural generation, and AI-assisted creative authoring in metaverse and media contexts; ETSI Domain AI + Data Mgmt addresses ML pipeline management, intelligent analytics, and AI-driven data governance; ETSI Domain AI + Governance encompasses ethics, explainability, bias detection, and regulatory compliance frameworks; and ETSI Domain AI + Human Interface covers conversational AI, gesture recognition, emotion detection, and intelligent user experience adaptation. Beyond these four direct sub-domain cross-markers, ETSI Domain AI also interfaces with cross-domain intersection concepts where AI meets other primary domains: ETSI Domain: Data Management + AI represents the data-management perspective on AI systems, while intersections with ETSI Domain Security and Privacy cover the security dimensions of AI system deployment. The domain concept thus functions as a multi-directional hub rather than a simple hierarchical root, acknowledging the deeply cross-cutting nature of AI technology across every domain of ETSI’s standardisation remit.

The concept emerged from ETSI’s recognition that AI-related standardisation activity had proliferated across many of its Technical Committees and Industry Specification Groups in an uncoordinated manner that risked fragmentation of vocabulary, duplication of scope, and gaps in coverage. By the mid-2020s, ETSI counted at least 14 Technical Committees (TC) and Industry Specification Groups (ISG) actively producing AI-relevant specifications. These ranged from ETSI ISG ENI’s (Experiential Networked Intelligence) cognitive network management and AI agent interface work — culminating in GS ENI 005 V4.1.1 (January 2026) — to ETSI TC SAI’s (Securing Artificial Intelligence, formerly ISG SAI) baseline cybersecurity requirements for AI models and systems expressed in ETSI EN 304 223 V2.1.1 (December 2025), and ETSI ISG MEC’s (Multi-access Edge Computing) edge computing AI enablers codified in GS MEC 003 V4.1.1 (May 2025). Additionally, ETSI ISG ZSM (Zero-touch Network and Service Management) defines AI/ML enablers for end-to-end network automation, TC Cyber addresses AI security dimensions outside SAI’s scope, and various programme committees in broadcast, transport, and consumer electronics produce specifications with AI components that must be classified within the taxonomy. ETSI Domain AI provides the common vocabulary node linking these diverse outputs into a coherent, navigable corpus that industry participants, regulators, and researchers can traverse systematically.

The functional importance of ETSI Domain AI as a classification concept has been substantially amplified by the progressive enforcement of the EU AI Act (Regulation 2024/1689). Under the Act, which entered force in August 2024 and became progressively enforceable from February 2025 (prohibited practices) through August 2025 (General Purpose AI obligations) to August 2026 (high-risk AI system obligations under Annex III), manufacturers and notified bodies must identify which technical standards apply to their AI systems. This identification is intrinsically domain-dependent: a high-risk AI system for medical diagnosis requires different harmonised standards from one for autonomous vehicle perception or for employment screening. ETSI Domain AI provides the indexing layer that makes this navigation tractable, mapping application contexts to applicable ETSI technical outputs and thence to the broader harmonised standards architecture being developed through CEN-CENELEC JTC 21. The harmonised standards process, managed jointly through JTC 21 and ETSI under a European Commission standardisation request accepted in May 2023, depends fundamentally on this domain classification structure to assign standardisation mandates to the correct technical bodies and to verify that the resulting standards cover all required application domains. ETSI RISE (Research, Innovation and Standardisation Ecosystem), the ETSI Board strategy group established to bridge the research community and standardisation work, uses ETSI Domain AI classifications to organise outreach to academic AI research groups and guide talent towards high-priority standardisation deliverables.

In the ontology underpinning this knowledge graph, ETSI Domain AI occupies a dual role: it functions as a formal OWL class (artificial-intelligence:ETSIDomainAI) representing a real-world institutional classification artefact, and simultaneously as a semantic hub linking AI technology domains (Computer Vision, Machine Learning, Natural Language Processing, Reinforcement Learning, Federated Learning) to governance and standards domains (AI Governance, EU AI Act, Harmonised Standards, AI Conformity Assessment). This bidirectional character — technology-facing on one side, governance-facing on the other — reflects ETSI’s institutional position as a standards body whose outputs must be technically rigorous while serving regulatory and market access functions. The class inherits from ComputationAndIntelligenceDomain (the broader AI research area) and is instantiated at the ApplicationLayer, reflecting that AI capabilities are expressed in application contexts rather than at the network infrastructure layer alone. This layer attribution also reflects a deliberate design choice in the ETSI taxonomy: AI is treated as an application-layer phenomenon — something that runs on top of network and infrastructure — even where AI is deeply integrated into network management functions (as in ENI’s cognitive management model), acknowledging that from a standards perspective, the AI intelligence layer can be separated from the transport and infrastructure layers for analytical purposes even if not in operational practice.

The historical lineage of the concept connects to ETSI’s earliest work on intelligent networks in the 1990s (IN-CS1 and IN-CS2 specifications), through the emergence of autonomic and cognitive networking research in the 2000s (ETSI work on NGN intelligent service management), to the first ETSI specifications explicitly addressing AI as a distinct concern (ISG ENI, established 2017; ISG SAI, established 2019; ISG MEC AI work items emerging from 2018 onwards). The transition from “intelligence” as an implicit property of advanced network services to “artificial intelligence” as an explicit classification category in ETSI’s taxonomy reflects the broader societal shift in AI discourse during the 2016-2020 period, when deep learning’s practical successes made AI a mainstream engineering concern rather than a specialised research area. The naming convention “ETSI Domain AI” follows ETSI’s broader taxonomy architecture in which each primary domain receives a canonical marker concept: ETSI Domain Data, ETSI Domain Security and Privacy, ETSI Domain Identity and Trust. This consistent naming enables automated tools to parse the taxonomy structure and perform domain-based queries across ETSI’s standards corpus, which by the mid-2020s ran to several thousand published specifications. The domain marker thus serves a machine-readable function in addition to its human-readable navigational role, making it an important element of ETSI’s digital standardisation infrastructure strategy.

The governance of ETSI Domain AI as a classification concept is distributed across multiple ETSI bodies. The ETSI Board Strategy Group RISE (Research, Innovation and Standardisation Ecosystem) has overall responsibility for ETSI’s AI standardisation strategy, including the domain taxonomy’s maintenance and evolution. Individual Technical Committees and ISGs (ENI, SAI/TC SAI, MEC, ZSM) maintain the specific work items classified under ETSI Domain AI, and their scope decisions directly affect what is in-scope for the domain node. ETSI’s work programme management team coordinates the domain tagging of new work items, ensuring consistent classification against the taxonomy’s controlled vocabulary. The European Commission’s standardisation request M/586 (accepted by CEN-CENELEC in September 2023) has created additional governance pressure: the harmonised standards programme requires ETSI’s domain taxonomy to be stable and coherent enough to serve as a compliance navigation tool for regulated manufacturers, elevating the stakes of taxonomy maintenance decisions. Any change to the ETSI Domain AI node’s scope definition or its controlled vocabulary now has potential legal consequences for manufacturers who have built compliance workflows on the taxonomy structure.

Components / Architecture

The ETSI Domain AI domain marker organises AI standardisation across several structural dimensions, each of which corresponds to a distinct layer of the taxonomy architecture:

  • Root Domain Class (Tier 1): ETSIDomainAI is the top-level OWL class in the AI branch of the ETSI taxonomy. All AI-related ETSI standards must map to this root or to one of its direct sub-classes. The root class carries the minimum necessary assertion that a work item’s primary character is defined by AI capability, leaving specific technology, application, and governance distinctions to lower-tier classification nodes. In OWL terms, ETSIDomainAI is declared as a named class with declared super-classes (ComputationAndIntelligenceDomain, AIResearchArea) and serves as the domain restriction in object property definitions that link AI work items to their sub-domain classifications.

  • Sub-Domain Cross-Markers (Tier 2) — direct hasPart relations from ETSIDomainAI:

    • ETSI Domain AI + Creative Media — generative AI, procedural content, AI-assisted authoring and rendering for media and metaverse contexts; in scope: ETSI ARF work on AI scene understanding, deepfake detection (GR SAI 011), AI-assisted content creation workflows
    • ETSI Domain AI + Data Mgmt — ML pipelines, data lakes, intelligent analytics, AI-driven data governance at scale; in scope: ENI data collection frameworks, MEC AI data plane specifications, ZSM ML-enabled operations data management
    • ETSI Domain AI + Governance — responsible AI, explainability, bias auditing, regulatory compliance, AI ethics frameworks; in scope: TC SAI security requirements, EU AI Act compliance specifications, explainability framework work items
    • ETSI Domain AI + Human Interface — conversational AI agents, voice interfaces, gesture/emotion recognition, adaptive UX; in scope: ETSI human factors specifications with AI components, AI-assisted accessibility standards

  • Secondary Cross-Domain Intersections — additional taxonomy nodes at the AI boundary:

  • Standards Producing Groups (requires relations from ETSIDomainAI):

    • ETSI ISG ENI — Experiential Networked Intelligence: cognitive network management, closed-loop automation, AI agent interfaces. Key outputs: GS ENI 005 V4.1.1 (System Architecture, January 2026), GR ENI 051 V4.1.1 (6G Native AI, February 2025), GR ENI 035 (AI Categorisation for IP Networks, December 2023), GS ENI 059 (AI Agent Interfaces, active 2025-2026)
    • ETSI TC SAI — Securing Artificial Intelligence (formerly ISG SAI): cybersecurity requirements, threat taxonomies, implementation guidance for AI systems. Key outputs: ETSI EN 304 223 V2.1.1 (Baseline Cybersecurity Requirements, December 2025), ETSI TR 104 065 V1.1.1 (May 2025), ETSI TR 104 128 V1.1.1 (May 2025), ETSI TS 104 223 V1.1.1 (April 2025)
    • ETSI ISG MEC — Multi-access Edge Computing with AI enablers: edge AI architecture, vertical API enablement, federation. Key outputs: GS MEC 003 V4.1.1 (Framework and Reference Architecture, May 2025), GS MEC 002 V4.1.1 (Requirements, June 2025), GR MEC 043 V4.1.1 (AI-enabled Vertical APIs, August 2025), GR MEC 047 (Distributed Computing)
    • ETSI ISG ZSM — Zero-touch Network and Service Management: AI/ML enablers for end-to-end automated network operations. ZSM defines the AI/ML capability layer for network-wide automation orchestrated across domains
    • TC Cyber (intersect with AI) — Cybersecurity specifications with AI components outside TC SAI scope
    • ARF (Augmented Reality Framework) — AI-driven scene understanding, spatial anchoring (GR ARF 012, November 2025), relevant to ETSI Domain AI + Creative Media

  • Technology Coverage (uses relations from ETSIDomainAI):

    • Supervised, unsupervised, semi-supervised, and self-supervised learning
    • Reinforcement Learning for autonomous network optimisation (primary ENI and ZSM use case)
    • Natural Language Processing for network management interfaces, log analysis, and ETSI Domain AI + Human Interface applications
    • Computer Vision for spatial computing (ARF), video analytics (MEC), and ETSI Domain AI + Creative Media applications
    • Federated Learning for privacy-preserving distributed AI in telecommunications (anticipated MEC and ENI work items)
    • AI agents and agentic systems with planning, memory, and tool-use capabilities (ENI GS ENI 059, 2025-2026)
    • Probabilistic inference, Bayesian networks, and expert systems (legacy intelligent network components within scope)

  • Regulatory Interface Layer:

    • Bidirectional mapping to EU AI Act risk classification categories (Annex I prohibited practices, Annex III high-risk system types)
    • Interface with CEN-CENELEC JTC 21 harmonised standards programme (approximately 35 standards, revised completion 2026)
    • Alignment with SC 42 AI foundational standards (ISO/IEC 22989 vocabulary, ISO/IEC 42001 management systems, ISO/IEC 23053 ML framework)
    • Connection to ITU-T AI standards (ITU-T Focus Group on AI for Health, ITU-T Study Group 13 recommendations on AI/ML in future networks)

  • Semantic Technology Interface:

    • OWL class hierarchy with DL-Lite (EL profile) axioms enabling automated classification reasoning

    • SKOS-aligned controlled vocabulary layer for human-readable term navigation

    • Linked Data interface enabling machine-readable cross-referencing with external standards databases

    • RDF representation supporting SPARQL-based standards gap analysis queries

      Use Cases / Major Families

      ETSI Domain AI functions as the classification anchor for an expansive range of real-world AI application scenarios encountered in ETSI’s standardisation work. Each of the following use case families maps to specific ETSI specifications, creating a traceable chain from abstract domain classification to concrete implementable standard:

      Intelligent Network Automation

      ETSI ISG ENI specifications address AI-driven network management where machine learning models adapt routing, resource allocation, and quality-of-service parameters in real time. ENI GS 005 V4.1.1 (January 2026) delivers a system architecture enabling closed-loop cognitive management with AI agent interfaces at the core, representing Release 4 of ENI’s technical programme. The closed-loop model — where AI continuously observes network state, infers optimal configuration, implements changes, and monitors outcomes — is the defining architectural pattern of intelligent network automation classified under ETSI Domain AI. GR ENI 035 (AI Categorisation for IP Networks, December 2023) provides a taxonomy of AI autonomicity levels applicable to specific network management domains, enabling classification of deployed AI systems by their level of autonomous decision-making authority — from fully manual through assisted, partial, conditional, high, and full autonomy. This multi-level autonomicity framework is directly referenced in ETSI Domain AI as a sub-classification dimension for network automation AI systems.

      Edge AI for Telecommunications and Verticals

      ETSI ISG MEC Phase 4 (from mid-April 2024) integrates AI/ML inference directly at edge computing nodes, enabling low-latency AI applications for augmented reality, autonomous vehicle coordination, industrial IoT, and smart city infrastructure. GR MEC 043 V4.1.1 (August 2025) specifically addresses the developer experience challenge by proposing abstracted APIs for vertical industry access to AI-enabled edge services, hiding the complexity of multi-access network heterogeneity behind standardised interfaces. GS MEC 003 V4.1.1 (May 2025) consolidates the reference architecture for MEC in a complex heterogeneous cloud-edge ecosystem, providing the structural model within which AI inference workloads are positioned. Edge AI classified under ETSI Domain AI includes both AI-as-a-service (AI inference exposed as API endpoints to applications) and AI-in-the-infrastructure (ML models embedded in MEC platform management functions), creating a classification dimension spanning application and infrastructure layers.

      Securing AI Models and Systems

      TC SAI (converted from ISG SAI around 2024) delivers ETSI EN 304 223 V2.1.1 (December 2025) — the first European standard and first global standard setting minimum cybersecurity requirements across the full AI lifecycle for all stakeholders in the AI supply chain. This specification, endorsed by UK NCSC and DSIT as the technical basis for the UK AI Cyber Security Code of Practice (published early 2025), addresses threats to AI systems across training, validation, deployment, and monitoring phases. ETSI TR 104 128 (May 2025) provides implementation guidance with cross-mappings to OWASP AI Exchange, MITRE ATLAS, and NIST Adversarial Attacks Taxonomy, making ETSI Domain AI security classifications interoperable with adjacent threat intelligence frameworks. ETSI TR 104 065 (May 2025) addresses specific threat categories including data poisoning, model inversion, membership inference, and adversarial example attacks — all classified as sub-categories within the AI security domain node.

      Augmented and Extended Reality AI Components

      Within ETSI’s Augmented Reality Framework (ARF) work, AI-driven components — including scene understanding, object recognition, semantic segmentation, spatial anchoring, and real-time environment mapping — are classified under ETSI Domain AI nodes in the taxonomy. ETSI GR ARF 012 V1.1.1 (November 2025) addresses AI requirements within augmented reality frameworks, establishing how AI-driven perception capabilities integrate with ARF architectural components. These AI-AR intersections map into ETSI Domain AI + Creative Media and ETSI Domain AI + Human Interface sub-domain nodes simultaneously, exemplifying the multi-domain tagging philosophy of the taxonomy.

      Generative AI in Creative and Media Industries

      The ETSI Domain AI + Creative Media sub-class captures generative content systems including text-to-image, text-to-video, and audio synthesis models deployed in media production workflows. GR SAI 011 addresses deepfake detection — the use of AI to identify manipulated multimedia identity representations — classifying both the threat (AI-generated deepfakes) and the countermeasure (AI-powered detection) within the ETSI Domain AI taxonomy. AI-assisted authoring tools for broadcast, gaming, immersive entertainment, and metaverse content creation are within scope, covering procedural generation of 3D assets, AI-driven animation, intelligent scriptwriting tools, and personalised content delivery systems.

      AI Governance, Ethics, and Regulatory Compliance

      The ETSI Domain AI + Governance sub-class maps to the proliferating compliance landscape under the EU AI Act, providing the domain classification layer connecting responsible AI practices (explainability, fairness, human oversight, transparency documentation) to applicable technical standards. This sub-domain anchors ETSI’s role in the harmonised standards programme under EU AI Act Article 40, which provides that AI systems compliant with harmonised standards are presumed to conform with the Act’s requirements covered by those standards. ETSI work items addressing AI conformity assessment frameworks, bias evaluation methodologies, algorithmic auditing procedures, and AI transparency reporting are all classified under this governance sub-domain.

      Conversational AI and Intelligent Human Interfaces

      The ETSI Domain AI + Human Interface sub-class covers voice assistants, dialogue systems, multimodal interaction systems, and AI-driven adaptive user experience components deployed in telecommunications and consumer technology contexts. ETSI human factors standards (ETSI EG 203 350 and related deliverables) addressing AI-driven accessibility — automatic captioning, speech enhancement, real-time translation, cognitive load adaptation — are classified under this sub-domain. The intersection of conversational AI with telecommunications network management (AI-driven network configuration through natural language interfaces) bridges the HCI and network management domains within the ETSI taxonomy.

      Formal Algorithm / Mathematical Treatment

      As a classification concept rather than an algorithm, ETSI Domain AI does not define a formal computational procedure. However, its role in the taxonomy can be formally characterised as a partition function over the space of AI-related standardisation artefacts. Given a set of ETSI work items W = {w₁, w₂, …, wₙ}, the ETSI Domain Taxonomy defines a classification function f: W → P(D) where D is the set of domain nodes {ETSIDomainAI, ETSIDomainData, ETSIDomainSecurityPrivacy, …} and P(D) denotes the power set (multi-domain tagging). A work item wᵢ is classified under ETSIDomainAI if and only if its primary technical scope includes at least one AI capability class C ∈ {MachineLearning, DeepLearning, NaturalLanguageProcessing, ComputerVision, ReinforcementLearning, ExpertSystem, FuzzyLogic, …}. The sub-domain cross-markers ETSIDomainAI{CreativeMedia, DataMgmt, Governance, HumanInterface} are defined as intersections: a work item wᵢ is classified under ETSIDomainAI+Governance if and only if f(wᵢ) ∩ {ETSIDomainAI, ETSIDomainGovernance} = {ETSIDomainAI, ETSIDomainGovernance}. The OWL formalisation in this knowledge graph expresses these classifications as ObjectSomeValuesFrom restrictions on the relevant object properties (hasPart, dependsOn, enables, etc.), enabling automated reasoning over the taxonomy structure using DL-based reasoners.

      Academic Context

      The intellectual foundations of ETSI Domain AI as a classification concept draw on several converging research traditions:

  • AI Taxonomy Research: Early taxonomies of AI application domains emerge from Nilsson (1998) and Russell & Norvig (2009), establishing the canonical partitioning of AI into perception, reasoning, learning, and action clusters that inform domain classification schemes. More recent taxonomies from the OECD (2019) and ITU (2019) on AI capabilities and application domains directly influenced ETSI’s own domain classification approach, with ETSI participating in ITU-T coordination activities on AI standardisation vocabulary.

  • Standardisation Studies: Research on how standards bodies organise technical knowledge (Egyedi & Mehos, 2012; Blind, 2004) provides theoretical grounding for why domain taxonomy structures are needed and how they function as coordination mechanisms across distributed technical communities. Büthe and Mattli (2011) on the geopolitics of international standards provides context for understanding ETSI’s relationships with ISO, IEEE, and ITU in the AI domain. The specific tension between ETSI’s distributed group model (many ISGs producing AI-relevant specifications) and the need for coherent vocabulary is a concrete instance of the coordination problem studied in standardisation economics.

  • AI Governance Frameworks: Work by Cihon (2019), Dafoe (2018), and the Oxford Internet Institute on AI governance architectures informs the design of the governance sub-domain concepts under ETSI Domain AI. The proliferation of AI ethics guidelines (studied by Jobin, Ienca & Vayena, 2019, who mapped 84 such documents) and their heterogeneous vocabulary created a driving need for standardised terminology that the ETSI taxonomy’s Controlled Vocabulary layer addresses.

  • Network Intelligence and ENI: The technical basis for ETSI ISG ENI draws on autonomic networking research (Strassner et al., 2009) and closed-loop cognitive management (ETSI GS ENI 001-010 series from 2017 onwards), positioning ENI’s AI domain as a specialisation of telecom network intelligence. The OODA (Observe-Orient-Decide-Act) loop model adapted for network cognitive management provides the conceptual architecture underlying ENI system specifications, making ENI’s domain classification within ETSI Domain AI a grounded technical decision rather than a purely administrative one.

  • Edge AI: Research on edge AI inference (Shi et al., 2016; Li et al., 2018) provides the technical grounding for ETSI ISG MEC’s AI work items, classified under ETSI Domain AI in the taxonomy. The distinction between cloud AI and edge AI is now a canonical division in the ML systems literature, and ETSI’s classification architecture reflects this by allowing both cloud-centric and edge-centric AI specifications to be classified under the same ETSIDomainAI root while differentiating their deployment context through additional property annotations.

  • Trustworthy AI: The EU High-Level Expert Group on AI (2019) Ethics Guidelines for Trustworthy AI, ENISA’s AI Threat Landscape reports (2020-2023), and the OECD AI Principles (2019) all inform both the governance sub-domain (ETSI Domain AI + Governance) and the SAI security work classified under ETSI Domain AI. The seven requirements for trustworthy AI identified by the EU HLEG — human agency and oversight, technical robustness and safety, privacy and data governance, transparency, diversity/non-discrimination/fairness, societal and environmental well-being, and accountability — map systematically to ETSI Domain AI sub-domain nodes and motivate the taxonomy’s cross-cutting governance markers.

  • Knowledge Representation for Standardisation: Formal ontology research by Smith et al. (2007) on biomedical ontologies and Noy & McGuinness (2001) on OWL ontology development provides the technical vocabulary for the semantic technology interface layer of the ETSI taxonomy. The application of Semantic Web technologies to standardisation corpora (linked open standards approaches) is an emerging field, and ETSI Domain AI’s design as a machine-readable taxonomy node anticipates a future where standards databases are navigable as linked data graphs.

    Standards and Context

    The ETSI Domain AI classification concept is contextualised within the following key standards and institutional frameworks:

  • ETSI ISG ENI: The Experiential Networked Intelligence group (established 2017) is the primary ETSI body producing AI specifications for cognitive network management. ENI Release 4 (GS ENI 005 V4.1.1, January 2026) establishes AI agent interface standards that represent ETSI’s most advanced published work on agentic AI in telecommunications.

  • ETSI TC SAI (formerly ISG SAI): The Securing Artificial Intelligence group (established 2019, converted to TC circa 2024) produces baseline security requirements for AI systems applicable regardless of application domain. ETSI EN 304 223 V2.1.1 (December 2025) is the flagship output, establishing minimum cybersecurity requirements across the AI lifecycle.

  • ETSI ISG MEC: The Multi-access Edge Computing group (established 2014) produces architecture and API specifications for AI-enabled edge computing. Phase 4 (from April 2024) focuses on heterogeneous cloud-edge-AI integration and developer-friendly vertical API abstractions.

  • ETSI ISG ZSM: Zero-touch Network and Service Management defines AI/ML enablers for end-to-end automated network service management, operating at the orchestration layer above individual network domains where ENI and MEC operate.

  • ISO/IEC JTC 1/SC 42: The primary international AI standards committee. ISO/IEC 22989 (AI vocabulary), ISO/IEC 42001 (AI management systems), and ISO/IEC 23053 (ML framework) provide the foundational vocabulary that ETSI’s domain taxonomy aligns with. SC 42’s annual plenary coordinates with ETSI through bilateral liaisons.

  • CEN-CENELEC JTC 21: The European joint technical committee on AI, commissioned by the European Commission in May 2023 to develop harmonised standards for the EU AI Act. JTC 21 works across approximately 35 standards projects, using ETSI’s domain taxonomy as a reference classification for assigning standards to application contexts.

  • EU AI Act (Regulation 2024/1689): The regulatory driver making ETSI Domain AI classifications commercially significant. Article 40 provides for harmonised standards creating presumption of conformity, while Annex III defines high-risk AI application domains that must map to ETSI taxonomy classifications.

  • ETSI GR SAI 011: Deepfake threat analysis — the use of AI for manipulating multimedia identity representations — classified at the intersection of ETSI Domain AI and security domains.

  • ETSI GR ARF 012 (November 2025): Augmented Reality Framework work on AI-driven scene understanding and spatial computing, classified under ETSI Domain AI + Creative Media.

  • NIST AI RMF: The US voluntary AI risk management framework; the ETSI Domain AI node contrasts-with this framework in being domain-classification-focused rather than risk-management-lifecycle-focused, though the two frameworks are increasingly used in conjunction by multinational organisations.

    Current Landscape (2026)

    As of June 2026, ETSI Domain AI is the most operationally significant classification category within the ETSI taxonomy, driven by the convergence of technological maturation and regulatory enforcement. Several developments mark the current period:

    Regulatory Enforcement Milestone

    The EU AI Act (Regulation 2024/1689) has reached its critical enforcement phase. Prohibited AI practices became applicable February 2025. GPAI model obligations entered force August 2025. The key deadline of August 2026 activates the full high-risk AI system regime under Annex III — meaning that manufacturers of AI systems in healthcare, biometric identification, critical infrastructure management, employment, education, law enforcement, and migration must demonstrate conformity with applicable requirements. This makes navigation via ETSI Domain AI classifications an active commercial workflow rather than a future aspiration. The ETSI taxonomy provides the domain classification layer linking a manufacturer’s AI system type to the applicable harmonised standards being produced by CEN-CENELEC JTC 21.

    Key Standards Milestones

    ETSI TC SAI published ETSI EN 304 223 V2.1.1 in December 2025 — the first European and first global standard setting minimum cybersecurity requirements across the full AI lifecycle for all supply chain stakeholders (developers, providers, deployers, and operators of AI systems). UK DSIT and NCSC co-developed this with TC SAI, making it the technical underpinning of the UK AI Cyber Security Code of Practice published in early 2025. ETSI ISG ENI published GS ENI 005 V4.1.1 in January 2026, with Release 4 work focusing on common ontologies and AI agent interfaces enabling true cross-domain automation from radio access to core and edge networks. GR ENI 051 V4.1.1 (February 2025) examines 6G native AI capabilities, foreshadowing the next major expansion of ETSI Domain AI scope.

    Edge Computing Evolution

    ETSI ISG MEC completed Phase 3 in mid-April 2024 and entered Phase 4, with GS MEC 003 V4.1.1 (May 2025) and GS MEC 002 V4.1.1 (June 2025) consolidating architecture for complex heterogeneous cloud-edge-AI ecosystems addressing MEC security enhancements, MEC federation, multi-domain and multi-tenancy slicing, and application slicing. GR MEC 043 (August 2025) specifically addresses AI-enabled vertical industry APIs, directly supporting ETSI Domain AI classifications for industrial and sector-specific edge AI deployments.

    Harmonised Standards Progress

    CEN-CENELEC JTC 21 progressed the harmonised standards programme, with prEN 18286 (AI Quality Management System for EU AI Act Regulatory Purposes) entering public enquiry on 30 October 2025 as the first harmonised standard in this series. Eight of approximately 35 planned work items had been published by early 2026. The overall programme, targeting completion by end of 2026 (revised from original April 2025 deadline), is considered the critical near-term deliverable for the ETSI Domain AI taxonomy’s practical utility as a compliance navigation tool.

    Institutional Evolution

    ETSI’s Research, Innovation and Standardisation Ecosystem (RISE) board strategy group was actively recruiting academic contributors to AI standardisation deliverables, broadening the epistemic base beyond incumbent telecoms operators. This institutional shift reflects the increasing importance of academic AI research to ETSI’s ability to produce technically credible specifications in areas like machine learning robustness, AI explainability, and agentic system safety — domains where telecoms industry expertise alone is insufficient.

    Benchmark Datasets and Evaluation

    ETSI Domain AI, as a classification concept, does not itself define or use benchmark datasets. However, the technical standards produced under the ETSI Domain AI umbrella reference several benchmark and test dataset contexts, and the sub-domain classification nodes inform which evaluation methodologies are applicable:

  • TC SAI Security Benchmarks: ETSI TR 104 128 (2025) references OWASP AI Exchange, MITRE ATLAS, and the NIST Adversarial Attacks Taxonomy as benchmark frameworks for evaluating AI system security. These provide the empirical grounding for TC SAI threat classification, which is annotated against ETSI Domain AI sub-domain nodes. The OWASP Top 10 for Large Language Model Applications, while not produced by ETSI, is cross-referenced in TC SAI governance domain work as a practical enumeration of AI application vulnerabilities.

  • ENI AI Performance Metrics: ETSI GR ENI 010 “Evaluation of Categories for AI Application to Networks” (declared stable as Release 4) and GR ENI 035 “Categorization of AI for IP Networks” (December 2023, Release 4) define evaluation criteria for AI autonomicity levels in network management, providing a benchmark-like classification framework for network AI capability maturity assessment. ENI 035 defines measurable criteria to score actual system benchmarks against defined autonomicity levels (from manual to fully autonomous), enabling classification of deployed AI network management systems within ETSI Domain AI.

  • MEC API Benchmarking: ETSI ISG MEC Phase 4 specifications (GR MEC 043 V4.1.1, August 2025) describe evaluation frameworks for AI-enabled edge APIs, implicitly defining performance benchmarks for AI inference latency, throughput, and availability at edge computing nodes. These benchmarks inform the classification of edge AI services within ETSI Domain AI at the performance dimension.

  • IEEE AI Standards Benchmarks: IEEE 2894-2024 (Architectural Framework for Explainable AI) defines benchmark structures for XAI evaluation that inform the ETSI Domain AI + Governance sub-domain and are cross-referenced in governance-relevant ETSI work items. IEEE standards on XAI performance metrics provide complementary evaluation vocabulary to ETSI’s own governance domain classifications.

  • EU AI Act Conformity Assessment Evaluation: Under Article 9 of the EU AI Act, high-risk AI systems must use relevant AI-specific standards for risk management, including performance benchmarking. ETSI domain-classified standards will inform which performance metrics are applicable to systems in each domain, effectively making domain classification the gateway to specific evaluation requirements.

    Sub-Domain Family: Four Intersection Nodes

    The four direct hasPart children of ETSI Domain AI constitute the primary sub-domain family. Each represents a distinct application sector where AI capabilities intersect with a non-AI primary domain. Understanding each is essential for accurate classification of AI standards within the ETSI taxonomy:

    ETSI Domain AI + Creative Media

    This intersection node captures the subset of ETSI specifications where AI capabilities are applied in creative production, media, entertainment, and immersive experience contexts. Primary ETSI work items classified here include:

  • GR SAI 011 — deepfake detection and manipulation of multimedia identity representations: classifies AI both as a threat vector (deepfakes) and as a countermeasure (AI-powered detection), demonstrating the bidirectional nature of AI in creative media security

  • ETSI GR ARF 012 — AI-driven scene understanding, object recognition, and spatial anchoring in augmented reality: captures AI as an enabling technology for immersive XR experiences

  • Generative AI specifications (forthcoming): as AI content generation tools proliferate in broadcasting, gaming, and metaverse contexts, new ETSI work items addressing AI-generated content quality, provenance, and interoperability are anticipated under this node The scope of this node is bounded by the creative output focus: AI used for logistics in a media company (supply chain AI) is not in scope; AI used to generate or analyse media content is in scope.

    ETSI Domain AI + Data Mgmt

    This intersection node covers AI specifications where data management, ML pipeline infrastructure, and intelligent analytics are the defining characteristics. Primary ETSI work items classified here include:

  • ETSI ISG ENI data collection and integration framework specifications — defining how network telemetry data is collected, pre-processed, and fed to ML models for network optimisation

  • ETSI ISG MEC data plane specifications — defining how AI inference workloads process data streams at edge computing nodes, including data format standards for AI model inputs/outputs

  • Forthcoming federated learning specifications — expected from ENI and MEC groups by 2027, covering privacy-preserving distributed ML training protocols The scope is bounded by the data management focus: AI algorithms themselves (the ML model architecture) are classified under the root ETSIDomainAI node; the data infrastructure enabling and surrounding those algorithms is classified here.

    ETSI Domain AI + Governance

    This intersection node covers AI governance, ethics, regulatory compliance, and responsible AI specifications — the most rapidly growing sub-domain as EU AI Act enforcement accelerates. Primary ETSI work items classified here include:

  • ETSI EN 304 223 V2.1.1 — baseline cybersecurity requirements for AI, representing the security governance dimension of responsible AI deployment

  • ETSI TC SAI threat taxonomy and countermeasure specifications — classifying AI threats and appropriate governance responses

  • Explainability framework work items — specifications addressing how AI systems must document and demonstrate their decision-making transparency

  • EU AI Act conformity assessment framework specifications — translating legal requirements into technical assessment procedures The scope includes both technical governance requirements (what must an AI system do to be trustworthy) and process governance requirements (what organisational procedures must surround AI system development and deployment).

    ETSI Domain AI + Human Interface

    This intersection node covers AI specifications where human-computer interaction, user experience, and human-AI collaboration are the defining characteristics. Primary ETSI work items classified here include:

  • ETSI human factors standards with AI components (EG series) — addressing AI-driven accessibility features (automatic captioning, speech enhancement, cognitive load adaptation)

  • Conversational AI specifications — defining interface standards for voice assistants and dialogue systems in telecommunications contexts

  • AI-driven adaptive UX specifications — covering AI systems that personalise user interfaces based on learned user behaviour models

  • Multi-modal interaction specifications — covering AI systems that combine voice, gesture, touch, and spatial inputs for immersive and XR interface contexts The scope is bounded by the human-facing character of the AI capability: AI used within a network management system (backend AI) is not in scope; AI used to create or adapt user-facing interfaces is in scope.

    ETSI Domain AI as a classification concept can be compared with adjacent AI categorisation systems:

  • vs NIST AI Taxonomy (NIST AI 100-1): NIST’s AI Risk Management Framework 1.0 (2023) and associated taxonomies classify AI by technical characteristics (supervised/unsupervised/reinforcement learning), application domains, and lifecycle phases. ETSI Domain AI focuses specifically on domain classification for standards navigation, without the lifecycle and risk management dimensions of NIST AI 100-1. The two frameworks are complementary: ETSI Domain AI for standards navigation, NIST AI RMF for risk management. US-EU aligned organisations use both.

  • vs EU AI Act Risk Classification (Annex III): The EU AI Act’s Annex III enumerates high-risk AI application sectors, which overlap substantially with ETSI Domain AI sub-domain nodes. However, the EU AI Act’s classification is risk-based (only certain applications in each sector are high-risk) while ETSI Domain AI is coverage-based (all AI applications in the domain are within scope). The ETSI taxonomy provides the bridge: from EU AI Act risk classification → ETSI domain node → applicable ETSI technical standards.

  • vs ISO/IEC 42001 (AI Management System): ISO/IEC 42001 classifies AI activities by their position in an AI lifecycle (design, development, deployment, operation, retirement). ETSI Domain AI classifies AI by application domain. These are orthogonal classification dimensions. A medical AI system, classified under healthcare in ETSI Domain AI, can simultaneously be in the deployment phase of ISO/IEC 42001 lifecycle. The two frameworks address different questions and are used together in comprehensive AI governance programmes.

  • vs IEEE AI Ethics Standards: IEEE’s portfolio of AI ethics standards (IEEE P7000 series) classifies AI concerns by ethical dimension (privacy, transparency, algorithmic bias, etc.). ETSI Domain AI classifies AI by application domain. IEEE P7000 standards can be relevant across multiple ETSI Domain AI sub-domain nodes, and the ETSI Domain AI + Governance node is the primary intersection point where IEEE ethics standards and ETSI domain classification converge.

  • vs OECD AI Classification: The OECD AI Policy Observatory classifies AI systems by technical characteristics (ML, knowledge-based, hybrid) and functional capabilities (recognition, generation, recommendation, etc.). ETSI Domain AI uses application domain as the primary classification axis, which is orthogonal to the OECD’s technical-capability axis. Cross-referencing the two classification systems enables more nuanced AI system characterisation.

    Formal Algorithm / Mathematical Treatment

    As a classification and navigation concept rather than a computational algorithm, ETSI Domain AI does not define a formal procedure. However, its role can be characterised formally using set theory and ontology constructs:

    Classification function: Given a universe of ETSI work items W and a set of domain nodes D, the ETSI taxonomy defines a classification relation R ⊆ W × D such that (wᵢ, dⱼ) ∈ R means work item wᵢ is classified under domain node dⱼ. For ETSI Domain AI, the relevant sub-relation is R_AI = {(wᵢ, ETSIDomainAI) : (wᵢ, ETSIDomainAI) ∈ R}, the set of all work items classified under ETSI Domain AI.

    Sub-domain inheritance: For each sub-domain intersection node dₛ (e.g., ETSIDomainAIGovernance), R_AI_s = {(wᵢ, dₛ) : (wᵢ, dₛ) ∈ R} ⊆ R_AI — every work item classified under a sub-domain intersection is also classified under the parent ETSIDomainAI node by inheritance.

    Standards gap detection: For a regulatory requirement set REQ = {r₁, r₂, …, rₖ} derived from the EU AI Act, and for each domain node dⱼ, the gap set Gap(dⱼ) = {rᵢ ∈ REQ : ¬∃wₖ s.t. (wₖ, dⱼ) ∈ R ∧ wₖ_covers(rᵢ)} identifies requirements not covered by any existing standard in domain dⱼ. Gap analysis systematically computes Gap(dⱼ) across all domain nodes, identifying where new standardisation work is needed.

    OWL formalisation: In the knowledge graph, ETSIDomainAI is expressed as an OWL named class with the following declared axioms (expressed in Manchester Syntax): Class: ai:ETSIDomainAI SubClassOf: ai:ComputationAndIntelligenceDomain, ai:AIResearchArea; hasPart some ai:ETSIDomainAICreativeMedia; hasPart some ai:ETSIDomainAIDataMgmt; hasPart some ai:ETSIDomainAIGovernance; hasPart some ai:ETSIDomainAIHumanInterface; partOf some ai:ETSIDomainTaxonomy; enables some ai:AIServiceClassification; enables some ai:AIConformityAssessment; standardizedBy some ai:ETSI.

    Key Terminology

  • Domain Marker: In the ETSI taxonomy architecture, a domain marker is a classification concept that identifies the primary technical domain of a work item without constituting a specification in itself. ETSIDomainAI is a domain marker for the AI domain.

  • Cross-Domain Intersection Marker: A taxonomy node representing the intersection of two or more primary domain concepts, used to classify work items with multi-domain scope. The four ETSI Domain AI sub-nodes (Creative Media, Data Mgmt, Governance, Human Interface) are cross-domain intersection markers.

  • Controlled Vocabulary: A standardised set of terms with canonical definitions, used to ensure consistent terminology across ETSI technical committee outputs. The ETSI taxonomy’s controlled vocabulary layer for AI draws on ISO/IEC 22989 terminology.

  • ISG (Industry Specification Group): A flexible ETSI body open to organisations outside formal ETSI membership, used to develop specifications more rapidly than the traditional TC route. ISG ENI, ISG MEC, and ISG ZSM are major AI-relevant ISGs.

  • TC (Technical Committee): A formal ETSI body with a defined scope for ongoing standardisation work. TC SAI is the main AI-security-focused TC; TC Cyber addresses broader cybersecurity with AI components.

  • Harmonised Standard: A European standard (EN) developed by a European Standards Organisation (ETSI, CEN, or CENELEC) under a European Commission standardisation request, whose application creates a presumption of conformity with the legal requirements it covers. ETSI EN 304 223 is an example relevant to ETSI Domain AI.

  • GPAI (General Purpose AI): A category defined in the EU AI Act for AI models with broad applicability across many tasks. GPAI obligations became enforceable from August 2025, directly relevant to standards classified under ETSI Domain AI.

    UK Context

    The United Kingdom maintains significant and multifaceted engagement with ETSI Domain AI-related standardisation across government, industry, and academia, despite no longer being an EU member state:

  • UK NCSC and DSIT: The National Cyber Security Centre and Department for Science, Innovation and Technology jointly co-developed the ETSI TC SAI baseline cybersecurity standard (ETSI EN 304 223 V2.1.1), released in December 2025. This collaboration made the UK government a direct shaper of content that sits within the ETSI Domain AI classification node, specifically the security dimensions of the AI domain. The UK AI Cyber Security Code of Practice (published early 2025) directly references this standard, creating a UK regulatory-standards pathway mediated through ETSI Domain AI classifications. The NCSC’s Annual Review 2025 chapter on AI highlights this standard as a key output of UK-led international AI security standardisation, signalling the NCSC’s ongoing engagement with ETSI TC SAI as a priority workstream.

  • BT Group Research (Adastral Park, Ipswich/Suffolk): BT chairs sessions on AI security within TC Cyber/TC SAI leadership at ETSI, reflecting BT’s long-standing position as one of ETSI’s most active members and its strategic interest in AI-driven network intelligence as a commercial differentiator. BT Research at Adastral Park — the UK’s largest industrial research campus outside London — has historically contributed to ETSI ISG ENI requirements, with BT’s network operations providing the operational use case grounding for ENI cognitive management specifications. BT’s 2025 commercial AI deployments across its consumer (EE) and enterprise networks in the UK directly instantiate the use cases classified under ETSI Domain AI.

  • Vodafone UK: As a major ETSI member and pan-European mobile operator, Vodafone contributes to ETSI ISG MEC and ISG ENI AI work items. Vodafone’s UK 5G network deployments in major English cities, and its private network AI applications for enterprise customers in the industrial Midlands and North of England, are the commercial context in which ETSI MEC and ENI specifications classified under ETSI Domain AI are deployed. Vodafone’s IPXI programme (Intelligent Programmable eXtensible Infrastructure) depends on AI/ML capabilities that map directly to ETSI Domain AI-classified specifications.

  • Arm Holdings (Cambridge): Cambridge-based Arm, whose processor architectures underpin approximately 99% of mobile device AI inferencing and the majority of edge computing hardware, participates in AI standardisation activities that intersect with ETSI Domain AI-classified MEC and edge AI specifications. Arm’s Machine Learning Group (formerly Ethos NPU team) contributes to standardisation bodies, and Arm’s architecture defines the hardware baseline assumed by many ETSI ISG MEC edge AI specifications. As a UK-headquartered semiconductor IP company of global significance, Arm’s participation in ETSI standards activities connects UK deep tech to global AI standardisation.

  • UKTIN (UK Technology Innovation Network): The UK Technology Innovation Network tracks and disseminates ETSI MEC Phase 4 developments to UK industry, translating ETSI Domain AI-classified standards into industrial strategy context. UKTIN reporting on ETSI MEC Phase 4 specifications (2024-2025) reflects UK government interest in edge AI infrastructure as a component of the UK’s National AI Strategy.

  • BSI (British Standards Institution): As the UK national standards body and ETSI member, BSI participates in ETSI Technical Committees and ISGs that produce specifications classified under ETSI Domain AI. BSI also represents the UK in CEN-CENELEC JTC 21, maintaining UK influence over harmonised standards mapped against ETSI taxonomy domain classifications.

  • Alan Turing Institute (London): The Turing Institute’s AI governance, safety, and explainability research directly informs the conceptual vocabulary of ETSI Domain AI + Governance domain nodes. Turing researchers participate in EU and UK AI policy processes that shape the regulatory requirements expressed through the taxonomy.

  • Northern England Industrial Context: In the Manchester-Leeds-Sheffield-Newcastle industrial corridor, AI deployment in manufacturing (smart factories, predictive maintenance at steel and chemical plants in the Humber), healthcare (NHS diagnostic AI across Yorkshire and Humber Integrated Care Systems), and transport (autonomous logistics on the A1/M1 corridor, Transpennine rail optimisation) represents the industrial ground truth that ETSI Domain AI-classified standards are designed to address. The 5G private network deployments enabled by ETSI ISG MEC specifications (classified under ETSI Domain AI) are being actively piloted at Sheffield Advanced Manufacturing Research Centre, Manchester Airport’s industrial campus, and the Siemens factory in Goole. These deployments instantiate AI service classification scenarios that the ETSI Domain AI taxonomy is designed to categorise.

    Future Directions (2026-2030)

    Several trajectories will shape ETSI Domain AI as a classification concept over the medium term, driven by technological evolution, regulatory developments, and institutional changes within ETSI and the broader European standardisation ecosystem:

  • 6G Native AI Integration: ETSI ISG ENI’s work on 6G native AI (GR ENI 051 V4.1.1, February 2025) foreshadows a fundamental shift in how AI is classified within the ETSI taxonomy. In 6G architectures, AI is not an application-layer service running on top of network infrastructure but rather an integral component of the network fabric itself — AI-native radio resource management, AI-native transport protocol optimisation, and AI-native core network function orchestration. The current ETSI Domain AI architecture, which instantiates AI at the ApplicationLayer, will need structural revision to capture these cross-layer AI capabilities. New sub-domain concepts capturing AI-native radio (intersection of ETSI Domain AI with radio access standards), AI-native transport (intersection with transport protocols), and AI-native core (intersection with network function virtualisation) are anticipated by 2027-2028. This evolution will also require alignment with 3GPP’s native AI work items in Release 19 and beyond.

  • AI Agent Standards and Multi-Agent Systems: ETSI ISG ENI’s pioneering GS ENI 059 (AI Agent Interfaces, 2025-2026) represents the leading edge of ETSI Domain AI’s expansion into agentic AI paradigms. As autonomous AI agents — capable of negotiating, collaborating, planning across time horizons, and executing complex multi-step tasks without human intervention — proliferate in network management, enterprise automation, and critical infrastructure contexts, ETSI Domain AI will need agent-specific sub-classification nodes. Multi-agent system coordination standards will require new taxonomy nodes sitting at the intersection of ETSI Domain AI and the existing network management domain markers. The distinction between reactive AI (current ML inference) and deliberative AI (agentic systems with planning and memory) is not currently captured in the ETSI taxonomy and will need to be introduced.

  • EU AI Act Harmonised Standards Completion and Maturation: The CEN-CENELEC JTC 21 programme (revised completion target end 2026, approximately eight months behind the original April 2025 deadline) will eventually produce approximately 35 harmonised standards for EU AI Act compliance. Once these are published in the Official Journal of the EU, they will carry a legal presumption of conformity, making navigation via ETSI Domain AI classifications a commercially critical compliance workflow. The maturation of this programme will establish the ETSI taxonomy’s domain classification structure as a de facto market access tool rather than merely an informational resource.

  • ISO/IEC 42001 Alignment and Convergence: The international AI management system standard ISO/IEC 42001 (published December 2023) provides a certifiable PDCA framework for AI governance. As global certification uptake accelerates — particularly in the wake of EU AI Act enforcement — the ETSI Domain AI + Governance sub-domain will increasingly reference ISO/IEC 42001 as the implementation vehicle for governance requirements expressed in ETSI taxonomy terms. While these two frameworks currently contrast in scope (ETSI taxonomy is domain-classification-focused; ISO/IEC 42001 is management-system-lifecycle-focused), they are increasingly used together in compliance workflows, creating pressure for convergence in vocabulary and mutual reference that will be reflected in future taxonomy revisions.

  • Federated Learning and Privacy-Preserving AI Standards: As Federated Learning and differential privacy techniques mature for telecommunications network optimisation, industrial IoT, and healthcare AI use cases, the ETSI Domain AI + Data Mgmt sub-domain will require expanded sub-classification to cover privacy-preserving training paradigms that do not centralise data. ETSI ISG ENI and ISG MEC groups are expected to produce work items addressing federated network intelligence and privacy-preserving model training for edge deployments by 2027. These will introduce new vocabulary nodes and relationship types into the ETSI taxonomy, likely requiring coordination with SC 42’s forthcoming standards on federated learning (anticipated in the SC 42 work programme).

  • Multi-modal AI and Immersive Reality Integration: The convergence of AI with XR (extended reality), spatial computing, and immersive metaverse environments will drive expansion of ETSI Domain AI + Creative Media and ETSI Domain AI + Human Interface to cover multi-modal AI systems operating simultaneously across visual, auditory, haptic, and spatial modalities. ETSI ARF (Augmented Reality Framework) work items (GR ARF 012, November 2025) are already introducing AI-driven scene understanding and spatial anchoring as standard components of XR systems. By 2028, ETSI Domain AI’s creative media and human interface sub-nodes will need to incorporate multi-modal AI concept families that do not yet exist in the taxonomy.

  • Digital Twin and Simulation AI: The proliferation of digital twin technology for network management (Digital Twin Network, a concept being developed across multiple ETSI groups) will require ETSI Domain AI to accommodate AI systems that operate on simulation/twin environments rather than physical network infrastructure directly. The classification of simulation-based AI training and digital twin AI inference systems will introduce new taxonomy dimensions not currently addressed.

  • Post-Quantum AI Security Intersection: As post-quantum cryptography standardisation (NIST PQC standards, ETSI quantum-related work items) converges with AI security (TC SAI scope), new cross-domain intersection markers at the boundary of ETSI Domain AI and ETSI Domain Security and Privacy will be needed, capturing AI systems that must be both adversarially robust (against ML attacks) and quantum-resistant (against cryptographic attacks on their security mechanisms).

    Research & Literature

    1. ETSI GS ENI 005 V4.1.1 (2026-01). Experiential Networked Intelligence (ENI); ENI System Architecture. ETSI. https://www.etsi.org/deliver/etsi_gs/ENI/001_099/005/04.01.01_60/gs_ENI005v040101p.pdf — The flagship ENI system architecture standard in Release 4, introducing AI agent interfaces and common ontologies for cross-domain automation; primary source for ETSI Domain AI classification of cognitive network management specifications.
    2. ETSI GR ENI 051 V4.1.1 (2025-02). Experiential Networked Intelligence (ENI); Analysis and Study of 6G Native AI Capabilities. ETSI. https://www.etsi.org/deliver/etsi_gr/ENI/001_099/051/04.01.01_60/gr_ENI051v040101p.pdf — Group Report analysing requirements for 6G native AI, covering radio access, transport, and core network AI-native functions; key forward-looking document for ETSI Domain AI taxonomy evolution.
    3. ETSI EN 304 223 V2.1.1 (2025-12). Securing Artificial Intelligence (SAI); Baseline Cyber Security Requirements for AI Models and Systems. ETSI — The first European and first global baseline AI security standard; co-developed with UK NCSC/DSIT; underpins UK AI Cyber Security Code of Practice; classified under ETSI Domain AI ∩ Security and Privacy.
    4. ETSI TR 104 065 V1.1.1 (2025-05). Securing Artificial Intelligence (SAI). ETSI. https://www.etsi.org/deliver/etsi_tr/104000_104099/104065/01.01.01_60/tr_104065v010101p.pdf — Technical Report supporting the TC SAI cybersecurity requirements with implementation context.
    5. ETSI TR 104 128 V1.1.1 (2025-05). Securing Artificial Intelligence (SAI). ETSI. https://www.etsi.org/deliver/etsi_tr/104100_104199/104128/01.01.01_60/tr_104128v010101p.pdf — Implementation guidance for ETSI EN 304 223 with cross-mappings to OWASP AI Exchange, MITRE ATLAS, and NIST Adversarial Attacks Taxonomy.
    6. ETSI TS 104 223 V1.1.1 (2025-04). Securing Artificial Intelligence (SAI). ETSI. https://www.etsi.org/deliver/etsi_ts/104200_104299/104223/01.01.01_60/ts_104223v010101p.pdf — Technical Specification within the TC SAI portfolio.
    7. ETSI GS MEC 003 V4.1.1 (2025-05). Multi-access Edge Computing (MEC); Framework and Reference Architecture. ETSI. https://www.etsi.org/deliver/etsi_gs/mec/001_099/003/04.01.01_60/gs_mec003v040101p.pdf — Phase 4 reference architecture for MEC in heterogeneous cloud-edge-AI ecosystems; primary source for ETSI Domain AI classification of edge computing AI specifications.
    8. ETSI GS MEC 002 V4.1.1 (2025-06). Multi-access Edge Computing (MEC); Requirements. ETSI. https://www.etsi.org/deliver/etsi_gs/mec/001_099/002/04.01.01_60/gs_mec002v040101p.pdf — Phase 4 requirements document for MEC, encompassing AI/ML inference requirements at edge nodes.
    9. ETSI GR MEC 043 V4.1.1 (2025-08). Multi-access Edge Computing (MEC). ETSI. https://www.etsi.org/deliver/etsi_gr/MEC/001_099/043/04.01.01_60/gr_MEC043v040101p.pdf — Group Report on developer-friendly abstracted APIs for vertical industry access to AI-enabled edge services.
    10. ETSI GR ARF 012 V1.1.1 (2025-11). Augmented Reality Framework (ARF). ETSI. https://www.etsi.org/deliver/etsi_gr/ARF/001_099/012/01.01.01_60/gr_ARF012v010101p.pdf — ARF Group Report addressing AI requirements for augmented reality; relevant to ETSI Domain AI + Creative Media and + Human Interface nodes.
    11. ETSI White Paper No. 52. ETSI Activities in the Field of Artificial Intelligence. ETSI. https://www.etsi.org/images/files/ETSIWhitePapers/ETSI-WP52-ETSI-activities-in-the-field-of-AI.pdf — Comprehensive overview of ETSI’s distributed approach to AI standardisation across 14+ technical bodies.
    12. ETSI Work Programme 2024-2025. Advancing Standards. ETSI. https://www.etsi.org/images/files/WorkProgramme/ETSI-Work-Programme-2024-2025.pdf — ETSI’s biennial work programme documenting all active work items and priorities.
    13. European Commission (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council (EU AI Act). Official Journal of the European Union — Primary regulatory driver elevating the commercial importance of ETSI Domain AI classifications.
    14. CEN-CENELEC (2025). JTC 21 Artificial Intelligence Work Programme Dashboard. CEN-CENELEC Joint Technical Committee 21 — Updated work programme for harmonised standards development under EU AI Act.
    15. ISO/IEC 42001:2023. Information Technology — Artificial Intelligence — Management System. ISO/IEC JTC 1/SC 42 — World’s first certifiable AI management system standard; contrasts-with ETSI Domain AI in being lifecycle-focused rather than domain-classification-focused.
    16. ISO/IEC 22989:2022. Information Technology — Artificial Intelligence — Artificial Intelligence Concepts and Terminology. ISO/IEC JTC 1/SC 42 — Foundation vocabulary standard providing the canonical definitions of AI terms that ETSI Domain AI’s controlled vocabulary aligns with.
    17. NCSC (2025). New ETSI Standard Protects AI Systems from Evolving Cyber Threats. National Cyber Security Centre. https://www.ncsc.gov.uk/blog-post/new-etsi-standard-protects-ai-systems-from-evolving-cyber-threats — UK government endorsement of ETSI EN 304 223; demonstrates UK-ETSI institutional alignment in AI security domain.
    18. NCSC Annual Review (2025). Chapter 3: Keeping Pace with Evolving Technology — Artificial Intelligence. https://www.ncsc.gov.uk/collection/ncsc-annual-review-2025/ — UK national cybersecurity strategy context for AI security standards.
    19. UK DSIT (2025). AI Cyber Security Code of Practice. Department for Science, Innovation and Technology — UK regulatory instrument built on ETSI EN 304 223; demonstrates practical deployment of ETSI Domain AI-classified standards in UK national policy.
    20. ETSI Artificial Intelligence Conference (2026). ETSI Standardisation in Support of AI. ETSI Press Release. https://www.etsi.org/newsroom/press-releases/2325-etsi-artificial-intelligence-conference — Conference documentation summarising ETSI’s distributed AI standardisation approach.
    21. Strassner, J. et al. (2009). The Role of Context-Aware Policies and Ontologies in Autonomic Networking. International Journal of Network Management, 1(1) — Foundational theoretical basis for ETSI ISG ENI’s closed-loop AI network management model.
    22. Shi, W. et al. (2016). Edge Computing: Vision and Challenges. IEEE Internet of Things Journal, 3(5), 637-646 — Canonical reference for edge computing architecture; technical grounding for ETSI ISG MEC AI work items classified under ETSI Domain AI.
    23. Li, H. et al. (2018). Learning IoT in Edge: Deep Learning for the Internet of Things with Edge Computing. IEEE Network, 32(1), 96-101 — Key research on edge AI inference applicable to ETSI MEC AI domain classifications.
    24. EU High-Level Expert Group on AI (2019). Ethics Guidelines for Trustworthy AI. European Commission — Seven requirements for trustworthy AI; conceptual basis for ETSI Domain AI + Governance domain node vocabulary.
    25. ENISA (2023). AI Threat Landscape Report. European Union Agency for Cybersecurity — Systematic threat enumeration for AI systems; informs TC SAI domain classification of AI security threats.
    26. Dafoe, A. (2018). AI Governance: A Research Agenda. Future of Humanity Institute, University of Oxford — Research programme framing standardisation as an AI governance mechanism.
    27. Cihon, P. (2019). Standards for AI Governance: International Standards to Enable Global Coordination in AI Research and Development. Future of Humanity Institute — Analysis of international AI standards landscape; directly relevant to ETSI Domain AI’s role in global AI governance.
    28. Jobin, A., Ienca, M. & Vayena, E. (2019). The Global Landscape of AI Ethics Guidelines. Nature Machine Intelligence, 1, 389-399 — Empirical mapping of AI ethics principles across 84 documents; provides vocabulary context for ETSI Domain AI + Governance classification nodes.

    Metaverse and XR Context

    ETSI Domain AI has specific relevance to the metaverse and extended reality (XR) application context that is the primary framing of this knowledge graph. Within the ETSI taxonomy structure, several metaverse-relevant AI classification points exist:

  • AI Scene Understanding for XR: AI components enabling real-time scene understanding, semantic segmentation, and spatial mapping in augmented reality applications are classified under ETSI Domain AI and specifically under ETSI Domain AI + Creative Media via the ARF (Augmented Reality Framework) connection. ETSI GR ARF 012 (November 2025) is the key standards reference.

  • AI-Driven Human Presence in Metaverse: AI components enabling realistic avatar representation, emotion detection, gesture recognition, and adaptive presence management in immersive environments fall under ETSI Domain AI + Human Interface, linking to the InteractionDomain and Telecollaboration branches of the knowledge graph.

  • AI at the Edge for Low-Latency XR: Low-latency XR rendering and interaction depends on edge AI inference at MEC nodes — the precise capability addressed by ETSI ISG MEC’s AI-related specifications classified under ETSI Domain AI. The [2ms interaction latency requirement for immersive XR](ETSI GR MEC 043 context) makes edge AI inference (classified under ETSI Domain AI) a core XR enabling technology.

  • Generative AI for Metaverse Content: Procedural content generation, AI-driven world-building, AI character behaviour, and AI narrative generation for persistent virtual worlds all fall under ETSI Domain AI + Creative Media, connecting to the CreativeMediaDomain branch of the taxonomy.

  • AI Governance in Metaverse Contexts: AI systems operating in metaverse environments — behavioural moderation AI, identity verification AI, content recommendation AI, economic transaction AI — all require governance frameworks classified under ETSI Domain AI + Governance, with direct relevance to TrustAndGovernanceDomain nodes in this knowledge graph.

  • Domain Marker Function in Metaverse Ontology: Within the metaverse ontology context (this knowledge graph’s primary focus), ETSI Domain AI functions as the anchor point connecting ETSI’s standards taxonomy to the AI capabilities that underpin metaverse components. Any metaverse component whose defining characteristic is an AI capability — regardless of application domain (healthcare metaverse, industrial metaverse, entertainment metaverse) — maps to the ETSI Domain AI classification node as its standards compliance entry point.

    Relationship to Adjacent Concepts in This Knowledge Graph

    The following ontology nodes are directly referenced by or cross-reference ETSI Domain AI within this knowledge graph. Understanding these relationships enables accurate traversal of the taxonomy:

  • ETSI Domain Taxonomy: The parent classification framework within which ETSI Domain AI is the primary AI-domain root node. ETSI Domain Taxonomy defines the structural rules (multi-domain tagging, controlled vocabulary, standards mapping layer) that govern how ETSI Domain AI is used as a classification concept.

  • ETSI Domain AI + Creative Media: Direct hasPart child. Classifies AI specifications concerning generative content, procedural generation, AI-assisted authoring, deepfake detection, and creative industry AI tools. Cross-references Computer Vision, creative media domain nodes, and ARF specifications.

  • ETSI Domain AI + Data Mgmt: Direct hasPart child. Classifies AI specifications concerning ML pipelines, data lake management, intelligent analytics, AI-driven data governance, and AI data plane specifications. Cross-references Machine Learning Framework, Data Governance, InfrastructureDomain nodes.

  • ETSI Domain AI + Governance: Direct hasPart child. Classifies AI specifications concerning responsible AI, explainability, bias detection, regulatory compliance, and AI ethics frameworks. Cross-references AI Governance Framework, Blockchain, TrustAndGovernanceDomain nodes.

  • ETSI Domain AI + Human Interface: Direct hasPart child. Classifies AI specifications concerning conversational AI, voice interfaces, gesture recognition, emotion detection, and adaptive UX. Cross-references Telecollaboration, InteractionDomain nodes, and conversational AI classification.

  • ComputationAndIntelligenceDomain: Super-class of ETSI Domain AI (referenced as AI Research Area in the json-ld blocks). Represents the broader computational intelligence domain of which ETSI Domain AI is the ETSI-standardisation-relevant sub-domain.

  • ApplicationLayer: The architectural layer at which ETSI Domain AI is implemented, indicating that AI capabilities are modelled as application-layer phenomena in the ETSI conceptual model even where they interact with network infrastructure.

  • ETSI GS MEC: Key source document. The Multi-access Edge Computing group specifications are the primary cross-reference point between ETSI Domain AI and edge computing deployment contexts. MEC Phase 4 specifications represent the most active current instantiation of ETSI Domain AI-classified edge AI standards.

  • AI Service Classification: An enabled capability. ETSI Domain AI’s taxonomy role enables structured classification of AI services, supporting product catalogues, procurement frameworks, and standards navigation tools that classify AI services by domain.

  • Intelligence Layer Categorization: An enabled capability. The taxonomy enables categorisation of AI systems by their intelligence layer characteristics — from rule-based systems through ML-based inference to agentic AI — within each application domain.

  • Computer Vision: A uses/implements relationship. Computer vision capabilities (object detection, scene understanding, image classification, spatial analysis) are among the core AI technologies within ETSI Domain AI’s scope, spanning both ETSI Domain AI + Creative Media (for content generation/analysis) and ETSI Domain AI + Human Interface (for gesture recognition) and edge AI (for video analytics at MEC nodes).

  • AI Governance Framework and AI Governance: Supported concepts. ETSI Domain AI’s governance sub-domain (ETSI Domain AI + Governance) maps to AI governance frameworks, making the ETSI taxonomy a structural component of AI governance infrastructure for the European market.

  • EU AI Act: Regulatory context that has elevated ETSI Domain AI from an informational classification aid to a component of the market access compliance architecture. The AI Act’s Annex III application domain list maps onto ETSI Domain AI sub-domain nodes.

  • Federated Learning: A uses/related-to relationship. Federated learning — privacy-preserving distributed ML without centralising training data — is a key AI methodology within scope of ETSI Domain AI, particularly for ETSI Domain AI + Data Mgmt and edge AI deployment contexts where data governance constraints preclude centralised training.

  • Deep Learning: A core technology within ETSI Domain AI’s scope. Deep neural network methods (transformer architectures, CNNs, RNNs, diffusion models) are the dominant implementation paradigm for AI systems in all four sub-domain nodes of ETSI Domain AI.

  • Natural Language Processing: Core technology for ETSI Domain AI + Human Interface (conversational AI, dialogue systems) and ETSI Domain AI + Data Mgmt (log analysis, network management NLP interfaces). NLP capabilities are among the most rapidly evolving within ETSI Domain AI scope, driven by large language model advances.

  • Reinforcement Learning: Core technology for ETSI ISG ENI’s cognitive network management model, where RL enables closed-loop autonomous optimisation of network parameters. RL-based AI is a primary implementation paradigm for ETSI Domain AI specifications in the network automation domain.

    Key Identifiers and Provenance Metadata

  • Legacy Term ID: 20330 (from original metaverse component classification system)

  • Preferred Term: ETSI Domain AI

  • OWL IRI: urn:ngm:class:etsi-domain-ai

  • VisionFlow Page IRI: urn:visionflow:page:babbbc5e48be6ec89517066976a3c2d1e23b129b76c632478d3b7928a3f9ad6d

  • Quality Score (original): 0.35 (stub quality, now enriched)

  • Inference Rule Applied: R5DomainRootFallback (original classification rule)

  • ETSI ISG ENI Portal: https://www.etsi.org/technical-groups/eni/

  • ETSI TC SAI Portal: https://www.etsi.org/technical-groups/sai/

  • ETSI ISG MEC Portal: https://www.etsi.org/technical-groups/mec/

  • ETSI AI Technologies Page: https://www.etsi.org/technologies/artificial-intelligence-automation/

  • Domain Equivalence: sameAs urn:ngm:class:etsi-ai-research-area (in the ComputationAndIntelligenceDomain branch)

Provenance

Graph View

Backlinks