ARIA Terminology Disambiguation: WAI-ARIA vs. Agent-ARIA

Status of this document

This is a short disambiguation note, as a reference by anyone indexing, citing, or searching on the term "ARIA" across W3C and W3C-adjacent work.

1. Background

Two distinct efforts currently use the bare acronym ARIA within, or adjacent to, W3C Community Group activity:

1. WAI-ARIA — Accessible Rich Internet Applications, a vocabulary of roles, states, and properties maintained by the Accessibility Guidelines Working Group under the Web Accessibility Initiative. It defines how dynamic web content and custom UI widgets expose semantics to assistive technologies. Current version: WAI-ARIA 1.3, Working Draft.
2. Agent-ARIA — Agent Registry for Identity and Authorization, a protocol stewarded by TrustLayer Foundation A.C. (a Mexican nonprofit constituted March 2026) that issues cryptographically signed, DID-anchored identity credentials to AI agents crossing organisational boundaries. It underlies the newly formed public-agent-identity W3C Community Group and is published under CC-BY-4.0, with a live registry at registry.aria.bar.

2. Comparison

3. Relation to other agent trust/interoperability Community Groups

Agent-ARIA does not sit in isolation. It is one of several concurrent, partly-overlapping efforts within the W3C Community Group ecosystem addressing AI agent identity, trust, and interoperability, formed in rapid succession through 2026:

• The Agent Identity Registry Protocol Community Group (proposed 22 April 2026, mailing list public-agent-identity) is Agent-ARIA's direct home: the group explicitly takes inspiration from TrustLayer Foundation's work without being constrained by it, and anticipates coordination with the W3C Credentials Community Group (CCG), the Decentralized Identity Foundation (DIF), the OpenID Foundation AIIM Community Group, and the IETF WIMSE Working Group.
• The AI Agent Protocol Community Group (proposed May 2025) is the broader umbrella under which agent identity models, security/privacy mechanisms, and protocol interoperability are scoped as named work items — several still pending community contribution.
• The Agent Trust Protocol (ATP) Community Group was formed at the direct invitation of the AI Agent Protocol CG chairs specifically to fill that CG's Security and Privacy modules, using pairwise DIDs, selective disclosure, and zero-knowledge proofs — a scope that substantially overlaps Agent-ARIA's trust-level and credential model and is a likely near-term convergence or duplication point.
• The Agent Declaration and Assurance Community Group (ADACG) addresses a related but distinct problem: cryptographic, verifiable declaration of an agent's identity, operational boundaries, and accountable party, rather than registry-based credential issuance.
• The AI Agent Memory Interoperability Community Group (proposed May 2026) is adjacent rather than overlapping — it reuses the same identity-binding primitives (post-quantum signatures, ML-DSA-65/FIPS-204) but applies them to portable agent memory rather than agent identity/authorization.

The practical implication for CGs is that at least three groups (Agent Identity Registry Protocol, ATP, and ADACG) are independently producing identity/trust specifications with visible scope overlap. This is a live standardisation-fragmentation risk worth tracking in the AIKR SDO inventory and landscape board

4. Vocabulary:

Agent-ARIA is not introducing a novel term vocabulary from first principles. It is built directly on two existing W3C Recommendation-track vocabularies that already supply the core primitives it needs:

• W3C DID Core — supplies DID, DID Document, verificationMethod, and service as the identity-anchoring layer.
• W3C Verifiable Credentials Data Model v2.0 — supplies issuer, holder, verifier, credentialSubject, and proof as the credential-exchange layer.

Agent-ARIA's own contribution sits on top of these as a thin, domain-specific extension: trust-level tiers (L0–L3, mapped to NIST SP 800-63-4 Identity Assurance Levels), the Agent Trust Policy format for declaring and evaluating intended actions, and the trust-ledger/revocation model. These are the terms that are genuinely new and not yet covered by DID Core or the VC Data Model.

Recommendation: Map/ align with is a short crosswalk note mapping Agent-ARIA's L0–L3 trust levels, Agent Trust Policy fields, and ledger/ revocation terms onto the existing DID Core and VC Data Model vocabulary, plus AIAO's ontic categories where agent accountability and delegation constructs apply. See Vocabulary Crosswalk: Agent-ARIA to DID Core, Verifiable Credentials, and AIAO for the term-by-term mapping. This positions AIKR as an interoperability bridge across the Agent Identity Registry Protocol, ATP, and ADACG groups rather than a fourth competing vocabulary source.

5. Recommendation for disambiguated usage

CG materials, indexes, and keyword matrices, this note recommends:

• Use WAI-ARIA (never a bare "ARIA") when referring to the accessibility vocabulary.
• Use Agent-ARIA or the full protocol name when referring to TrustLayer Foundation's agent identity/authorization work, to avoid collision with WAI-ARIA in search and citation.
• Where a bare "ARIA" appears in third-party source material, resolve it against context before indexing rather than assuming either sense by default.