The AI landscape is dense, chaotic, and evolving exponentially. In this context, terms, concepts, and initiatives overlap and are frequently ambiguous, making it more difficult to build the
shared understanding necessary to achieve interoperability and trust in agentic systems.The AI Knowledge Representation Community Group (AI KR CG) seeks to reduce the gaps in shared concepts and semantics. It monitors the evolution of AI across multiple dimensions — identifying, mapping, and analysing the primitives emerging from research and related standardization efforts. Using canonical knowledge-representation techniques, the AI KR CG builds shared, explicit representations — categories, taxonomies, and vocabularies — that support alignment, consistency, and model integrity across disparate initiatives, giving stakeholders common terms of reference for a fast-moving field.Through work on taxonomies, vocabularies, a standards inventory, gap analysis, and more, the Community Group hopes to help individuals and organisations stay up to date with developments in the field and achieve a deeper understanding of the ongoing evolution of the agentic-AI landscape
This repository holds shared notes and memos aimed at informing AI KR CG participants about the state of the art, through the representation and definition of key terms and concepts in the fast-moving agentic-AI landscape.
The content of these files is open for discussion, evaluation, and refinement. Because the space is emergent and the work is iterative, some of the older files may contain duplicate, missing, or redundant information that has since been merged into the newer files below. Nothing here is a W3C Standard or on the W3C Standards Track.
Participants are welcome to contact the chair, discuss via the public mailing list, or suggest how to develop these resources further.
Drafts arising from the trust-layer thread on public-aikr (opened 3 August 2026). All files live in the trust folder and are open for discussion; nothing here has been agreed by the group.
-
Use-case cards drawn from agent incidents and scenarios, a shared glossary, a trace from each weak point to proposed fixes and KRT terms, a coverage grid of contributors’ proposals, and plain-English summaries of the latest list exchanges. Proposed for discussion at the weekly calls from 21 September 2026.
-
Classes, properties, controlled values and SHACL validation rules R0–R11 for recording trust decisions involving AI agents. Files:
krt.ttl (ontology),
krt-shapes.ttl (shapes),
krt-example.ttl (one passing and one failing record).
-
Draft technical note on resolution dependency in agent identity, addressed to the W3C groups working on agent trust, with a survey of related efforts outside W3C and annexes on Base L2, MolTrust and the Agent Identity Protocol. Consultation opened 2 August 2026.
-
Full text of §4.1 of the trust-layer note: why canonical bytes do not establish that two artefacts concern the same subject, the three failure classes (Unicode normalization, large integers, divergent declared semantics), and what a profile must state.
-
Findings from queries against the ARIA public verification surface (8–9 September 2026), set against the wider field of DNS-anchored agent identity proposals. Prepared for the TrustLayer Foundation and offered to the AI KR CG for discussion; open for correction.
-
The DNS root key change of 11 October 2026, how the Credentials CG is discussing DNSSEC-anchored DIDs, implications for the open and agentic web, open questions on risks to everyday users, and links to the related TPAC 2026 issues.
-
A focused crosswalk of the identity/trust-specific W3C groups (Agent Trust Protocol CG, Agent Identity Registry Protocol CG, W3C Credentials CG, and related DID/VC work) converging on overlapping agent-identity vocabulary.
-
Disambiguates the accessibility WAI-ARIA vocabulary from the unrelated “Agent-ARIA” identity/authorization proposal circulating in agent-identity discussions — a naming collision worth flagging early. See also the companion
plain-language explainer on agent identity and authorization.
Earlier and supporting notes. Several of these fed into the three core resources above; some information they contain has since been merged or superseded. Retained for reference and history.
-
A crosswalk of current W3C Working Groups, Community Groups, and the Web & AI Interest Group whose scope touches AI or agentic AI — home page, repo, spec links, and main vocabulary terms for each, with notes flagging where groups are converging on overlapping terminology independently.
-
Summary, ecosystem alignment notes and a SKOS-style glossary built from the draft Commission guidelines on high-risk classification under Article 6 of the AI Act; prepared for the CG’s response to the July 2026 targeted consultation.
-
A ReSpec overview of the group’s purpose and how the documents in this repository fit together.
-
An interactive graph of the repository documents and their links to neighbouring W3C and IETF standards.
-
An interactive board placing each effort by framework layer (representation → trust) and openness.
-
The earlier standards inventory; largely consolidated into the Standards Inventory & Gap Analysis in the core resources above.
-
The June working draft of the vocabulary as tables by domain; retired and superseded by the Agentic AI Vocabulary in the core resources above.
-
A use-case alignment (not a new ontology) between the IDL Agent Ontology and the HMAS ontology stack.
-
The open e-commerce vocabulary stack and what it lacks for autonomous agents; input to the W3C/GS1 workshop.
-
A model-card framework for AI/ML systems used in digital and web forensics, with a JSON schema and reference implementation.
-
Notes on the proposed WebMCP browser API and its relationship to W3C Web Machine Learning incubation work.
-
A baseline vocabulary for agentic voice interaction, organised under the CG’s eight knowledge domains, with appendices on implementation-layer terms from open-source frameworks and on terms from the W3C Smart Voice Agents Workshop report. Opened for contributions 2 June 2026.
-
Requirements and design principles for voice AI systems, with analysis of documented deployment failures. See the
document-set guide for how the voice drafts relate.
-
Extended version with new voice-AI interaction-type classifications; from the W3C Smart Voice Agents Workshop.
-
A practical introduction to the W3C Solid protocol for content creators, AI agent developers, and users.
-
Feedback posted to the AI Agents Protocols CG (March 2026).
-
Guidance on authoring specifications readable and consumable by AI agents and automated tools.
-
Principles for accessibility and sustainability in AI code assistants.
-
Quality and accessibility principles for code generated by AI systems.
-
A survey of efforts intersecting the Agent-to-Web Framework (A2WF).
-
An explainer on robots.txt and content-usage signals in the context of AI agents.
