Digital wallets promise greater control over identity. They can reduce repeated identity checks, support the selective release of credentials, and let a person prove a necessary fact without disclosing everything they know about themselves. But holding a credential in an app is not the same as having durable control over it.
What happens when the phone is lost, the vendor discontinues the wallet, an issuer disappears, or the service requesting the credential is not who it claims to be? And when an AI agent has been given access through that wallet, can the principal recover or revoke the relationship after switching providers?
Two October developments bring those questions into sharper focus. On October 20, the FIDO Alliance announced public working documents for a credential-manager reference architecture and a verifier-identification framework, alongside a new advisory forum. On October 23, European Commission materials describing the 2026 EUDI Wallet Launchpad testing track emphasized practical interoperability among national wallets, issuers, relying parties and standards implementations.
Neither announcement means that all wallets are interchangeable today. Both suggest the conversation is moving beyond proof of concept toward operational trust between many organizations.
A trusted wallet needs a trusted verifier
Wallet discussions often concentrate on what a user can prove. Yet a user also needs to understand who is asking for evidence. A malicious site can imitate a government, bank or employer while asking for credentials that a legitimate party would never need.
FIDO's October 20 announcement describes a proposed framework for verifier identification. Its goal is to help credential managers present structured, verified information about the service requesting data. FIDO also describes a Credential Manager Reference Architecture for functional, security and interoperability expectations, intended to support a certification program. These were announced working documents, not a blanket assertion that every wallet has achieved certification.
This reciprocal-trust idea matters for AI agents. An agent can faithfully relay a credential request from a fraudulent service. If the wallet cannot establish who the requestor is, the agent's obedience may become a security liability.
MoonVault view: identity is not sovereign if the owner cannot tell who is asking, limit what is disclosed, and recover control when the relationship changes.
Europe is testing interaction between real ecosystems
The European Commission's EUDI Wallet Launchpad 2026 testing materials, updated October 23, describe the practical work of connecting national wallet teams, issuers, verifiers, test laboratories, conformity bodies and relying parties. The test scope includes OpenID for Verifiable Credential Issuance, OpenID for Verifiable Presentations, higher-assurance interoperability profiles and relevant standards for proximity interactions and trusted lists.
The distinction is important. Publishing a credential format does not guarantee that two independently built wallets, national identity systems or merchant services can transact safely together. Interoperability emerges through compatible protocols, trust registries, implementation testing, clear error handling and governance among parties that do not share a vendor.
The Commission's materials describe planned and coordinated testing activity, not independent proof that every scenario has passed. The credible signal is the emphasis on operational verification before broad adoption.
Interoperability is not the same as portability
Two wallets may support the same presentation protocol and still make migration painful. A credential may be bound to a device-held key. An issuer may require fresh enrollment. A user's recovery factors may be proprietary. Delegated permissions may live in an application database rather than an exportable, verifiable record.
There are at least four distinct questions:
- Can the credential be presented? Does the new wallet communicate with the verifier using supported protocols?
- Can the holder recover? What happens after loss of the original device or authenticator?
- Can the credential or relationship move? Can a trusted issuer rebind, reissue or recognize it without recreating the entire identity history?
- Can delegated authority be reconciled? After migration, which agents and services retain valid permissions, and who can revoke them?
Many systems will reasonably prevent raw private keys from being exported. That can be a security strength. Durable independence therefore may require secure reissuance, cryptographic key rotation, transferable policy records and a recovery process that does not weaken identity assurance. Sovereignty should not be confused with copying every secret to another app.
Agent-to-wallet designs show why the boundary matters
The October 17 individual Agent-to-Wallet Protocol draft describes one possible interface in which an agent passes credential requests to a wallet that retains credentials and performs the protected operations. It proposes constraints on agent-scoped access and discusses trusted counterparties, phishing, the origin of continuation URLs and attacks using manipulated agents.
The proposal is not an endorsed IETF standard. Its significance is the architectural split: the agent is a requester, while the wallet remains a controlled credential boundary. That distinction can outlast any particular agent vendor, but only if the surrounding identifiers, grant records, approval flows and recovery methods are interoperable.
Continuity becomes the missing test
Consider an owner who uses an AI assistant for household and business administration. A device is replaced. A company changes its identity provider. A trusted employee leaves. The owner becomes unavailable. The wallet application shuts down.
For each transition, the system should answer who still has authority, how evidence of prior approvals is preserved, how outdated grants are terminated, and how a new lawful principal or delegated steward can assume responsibility without inheriting unrestricted secrets.
These are separate from the ordinary task of proving age or employment status at a website. They require an explicit governance and recovery model, especially where rights persist longer than the device or software that originally expressed them.
MoonVault's long-horizon thesis is that identity, delegated authority, custody, revocation and continuity belong to one coherent control architecture. Whether a single platform should provide all those functions remains an open design and business question. Building them as independent interoperable services may ultimately be more resilient than concentrating everything in one proprietary vault.
The October FIDO and European testing efforts are encouraging because they push vendors toward shared requirements and verifiable behavior. The next standard of credibility should be stronger still: a person should be able to replace a device or provider without losing the ability to prove identity, terminate delegations and recover rightful control.
Earlier MoonVault analysis: why selective disclosure changes the privacy bargain.
Primary sources and evidence status
- FIDO Alliance, digital credentials and verifier identification announcement, October 20, 2026 (working documents and certification direction).
- European Commission, EUDI Wallet Launchpad 2026 testing page, updated October 23, 2026 (testing program, not universal conformance evidence).
- Agent-to-Wallet Protocol, revision 01, October 17, 2026 (individual Internet-Draft, not adopted standard).
- W3C Verifiable Credentials Data Model v2.0 (existing web standard, background).