eIDAS 2.0 wallets as identity infrastructure for European agentic commerce

The revised eIDAS Regulation requires every EU member state to make a conformant European Digital Identity Wallet available to any citizen or resident who requests one, with September 2026 as the operative compliance deadline. The mandate is broader than its predecessor: where the original eIDAS framework [11] addressed cross-border electronic identification and trust services at a relatively narrow technical layer, eIDAS 2.0 extends the scope to include qualified electronic attestations of attributes, a new category of trust service that governs how verified personal and organisational data moves across the single market.

The original eIDAS rollout offers an instructive precedent, though the structural differences matter. Member states deployed national eID systems to a directive deadline, and cross-border interoperability failures became measurable only after those systems were live; the most substantive technical refinements to notification procedures and interoperability nodes followed from that post-deployment evidence rather than from pre-deployment specification work [11][5]. That history supports the argument that deployment precedes specification maturity. The relevant structural difference is that the original framework involved human-in-the-loop authentication, where failure produced a declined login and no downstream consequence; agentic commerce involves autonomous execution, where a failure in an attested chain produces a completed transaction with unresolved liability.

Agentic commerce (transactions initiated or completed by autonomous software agents acting under delegated authority) was not a primary drafting concern when the Architecture and Reference Framework was published. The ARF's trust model addresses human wallet holders presenting credentials to relying parties; it does not contemplate a credential presentation chain in which the presenting actor is a software process operating within scope of a legal-person mandate. This gap sits between the framework's technical completeness and the commercial trajectory it is expected to serve. The regulatory text creates the obligation; the implementing architecture must now be extended, through delegated acts or standards, to close the distance between what the directive requires and what the deployed infrastructure can actually attest [14].

The European Digital Identity Wallet operates through a layered credential architecture. At its base, a Wallet Unit Attestation binds a specific wallet instance to a certified device and to the issuing wallet provider, establishing the hardware-rooted trust anchor from which all subsequent credential presentations derive their assurance. Attribute attestations (the signed claims about a holder's identity, qualifications, or entitlements) are issued by Qualified Electronic Attestation of Attributes providers, a class of Qualified Trust Service Provider recognised under eIDAS 2.0. Each attestation is encoded in either SD-JWT or mDOC format, both of which support selective disclosure: the holder can present a subset of attested attributes to a relying party without exposing the full credential [15].

The chain of trust flows in one direction. A QTSP signs an attribute attestation referencing the holder's verified identity. The holder's wallet stores that attestation and presents it, along with the Wallet Unit Attestation, to a relying party. The relying party verifies both signatures against the EU Trusted List, completing the chain from issuer to presentation. This architecture is coherent for natural-person holders; the cryptographic suites and signature formats employed, including ECDSA and the CAdES advanced electronic signature format specified in ETSI EN 319 122, are operationally established within European trust-service infrastructure.

The structural constraint appears at delegation. The ARF defines no credential type for expressing that a software agent holds authority granted by a natural person or legal entity; it assigns no assurance level to such a delegation claim; and it places no binding on how a relying party should resolve a delegated presentation back to the originating holder. These three absences are related: without a defined credential type, an assurance level cannot be assigned, and without an assurance level, a relying party has no conformant verification path. The SSI literature documents this individual-centric design limitation as a systemic gap across decentralised identity specifications, not one peculiar to EUDIW [6][12]. For agentic commerce, the absence is not peripheral: delegated authority is the transaction's first-order identity claim.

Merchant authentication architectures must be redesigned around attribute attestation, not session tokens. Under eIDAS 2.0, a relying-party platform accepting EUDIW credentials cannot treat the wallet presentation as a simple authentication event analogous to OAuth or a federated login. It must verify the Wallet Unit Attestation, the QTSP-issued attribute attestation, and the holder's selective disclosure proof as a compound artefact. This requires investment in EU Trusted List resolution, SD-JWT and mDOC parsing, and presentation-request protocol support, none of which is standard in existing European merchant stacks. The compliance deadline binds relying parties and member states alike [15][10].

Member-state attribute harmonisation creates a coordination dependency with no current resolution mechanism. Attribute attestations issued in one member state must be semantically interoperable with relying-party expectations in another. The EBSI cross-border pilot record demonstrates that attribute schema alignment does not emerge automatically from shared infrastructure: jurisdictions adopt different attribute definitions, encoding conventions, and revocation approaches, and reconciling these requires bilateral or multilateral governance work that precedes technical deployment [5]. Without a binding attribute schema registry or a published harmonisation schedule, cross-border commerce faces the prospect of per-corridor negotiation rather than single-market uniformity.

Relying-party platforms face a compressed timeline against a partially specified target on two fronts. The ARF does not define attribute semantics for non-human actors, and it provides no delegation credential type, assurance level, or relying-party resolution method for agentic transaction chains [6][12]. These are distinct gaps with distinct remediation paths: the attestation-semantics gap may be closed through a targeted implementing act extending existing QTSP obligations, whereas the delegation-chain gap requires a more fundamental ARF revision that assigns assurance levels to multi-hop authority claims. A platform investing in EUDIW integration before either gap is closed accepts specification risk on both fronts, and any commercial deployment of agentic transaction flows before that publication operates without an established liability-allocation framework for attested-chain failures.