Skip to content
IAM Insights

Explainers · Glossary

DID / Verifiable Credentials

DID · VC · verifiable credentials · Decentralized Identifiers

By Boston Identity · Explained · 5 min read

The short answer

Decentralized identifiers name entities and point to verification information. Verifiable credentials carry statements an issuer makes about someone or something. Together they let a holder present checkable evidence, while the receiver still decides whom to trust.

What it is
A standard
Defined by
W3C: VC Data Model 2.0 and DID Core 1.0 Recommendations
Who it's for
Teams evaluating portable credentials and agent identity
In our briefings
94By month of the source: 2025-12: 2, 2026-01: 1, 2026-02: 4, 2026-03: 4, 2026-04: 8, 2026-05: 4, 2026-06: 3, 2026-07: 2, 2026-08: 5, 2026-09: 5, 2026-10: 17, 2026-11: 00by month, last 12
01

What is DID / Verifiable Credentials?

A decentralized identifier, or DID, is an identifier such as did:example:123. Its method, the middle part of the name, determines how software finds its associated DID document. That document can describe verification keys and services. The subject named by a DID and the controller allowed to change its document can be different entities.S2 §1.1,S2 §1.3

A verifiable credential, or VC, carries an issuer's statements about a subject in a form whose authorship can be checked cryptographically. A degree, employment record or qualification can travel from an issuer to a holder and then to a verifier. The issuer, holder and verifier perform separate roles; the holder need not be the person the statements describe.S1 §1.1,S1 §1.2,S1 §2

These are separate W3C standards, often discussed together. VC Data Model 2.0 is a Recommendation dated 15 May 2025. DID Core 1.0 is a Recommendation dated 19 July 2022. Credentials do not depend on DIDs: the issuer must have a URL identifier, and a DID is one possible choice. DID Core also allows systems built on centralized or federated infrastructure, so the name does not guarantee a particular governance model.S1 §4.7,S1 §4.4,S2 §1

A DID is not a credential, and a credential does not need a DID. One names an entity and connects it to a document; the other carries an issuer's statements. Neither by itself settles whether a service should authorize an action.S2 §1.3,S1 §4.4,1

02

How does DID / Verifiable Credentials work?

Follow a university, a graduate and an employer. This example uses a DID for the issuer; other credential deployments can use different identifiers.S1 §1.1,S1 §4.7

Figure steps
  1. The issuer creates a credential
  2. The holder keeps the credential
  3. The holder presents the evidence
  4. The verifier resolves a DID when used
  5. The verifier checks proof and policy
Fig. 1

The graduate carries the credential between issuer and verifier; the employer checks its proof and makes its own trust decision.

Drawn from VC Data Model 2.0 and DID Core 1.0

Swipe for the full figure →

  1. The issuer creates a credential

    An issuer makes statements about a subject, such as a university recording a graduate's degree. The credential identifies its issuer and describes its subject. The issuer must secure conforming credentials with a supported securing mechanism. An embedded proof puts the proof inside the document; an enveloping proof wraps the document. Neither choice requires a DID.S1 §1.2,S1 §1.3,S1 §4.7,S1 §4.12

  2. The holder keeps the credential

    The issuer sends the credential to a holder, who stores it in a credential repository: software such as a wallet or protected storage vault. The holder may be the subject, but does not have to be. A parent holding a child's credential is the specification's example. Our walk-through uses a graduate holding their own degree credential.S1 §1.2,S1 §2

  3. The holder presents the evidence

    The holder generates a presentation from one or more credentials and shares it with a verifier, such as an employer checking a qualification. Some cryptographic schemes let the holder disclose only selected statements or prove a property without revealing the original credential. Those capabilities depend on the scheme; the word 'verifiable' alone does not promise them.S1 §1.1,S1 §2,S1 §4.13

  4. The verifier resolves a DID when used

    In this example the issuer uses a DID. A resolver applies that DID method's read operation to return a DID document. The document can identify public keys and the purposes for which they may be used. For a credential proof, the verifier checks whether the verification method belongs to the document's assertionMethod relationship. A key listed for authentication is not automatically a key for issuing credentials.S2 §7.1,S2 §5.3,S2 §5.3.2

  5. The verifier checks proof and policy

    Verification evaluates whether the credential or presentation is authentic and current. It checks conformance, the securing mechanism and, if present, credential status. The verifier then applies its own business rules before relying on the statements. A valid proof does not establish that the degree is true, that this issuer is worth trusting, or that the presenter should get access.S1 §2,S1 §1.1,S1 §5.1

03

Why does DID / Verifiable Credentials matter now?

Agent identity proposals are making these choices concrete. KYA-OS requires decentralized identifiers for agents and support for did:key and did:web.2 Its delegation credentials add rules beyond a general credential: each link must keep the same or narrower scope, and the verifier must check the whole chain.3 The same specification leaves open whose identifiers and issuers a receiver should accept.4

Other credential work takes a different route. The SD-JWT VC revision we read finds issuer keys through HTTPS metadata or certificate chains, leaves DID mechanisms to profiles and extensions, and explicitly distinguishes its format from the W3C data model.5 OpenID's credential-issuance specification uses OAuth and does not require DIDs either.6 We read those choices as a reason to ask separately which credential format a product supports and how it names its issuers.5,6

04

Where does it fit?

Compared withHow DID / Verifiable Credentials differs
KYA-OSKYA-OS applies DIDs and W3C credentials to agents and defines delegation-chain checks. The core credential model leaves authorization to an accompanying framework.3,1
SD-JWTSelective disclosure lets a holder share only some statements. W3C credentials can use selective-disclosure schemes, but support depends on the securing mechanism; it is not guaranteed for every credential.S1 §2,S1 §4.13
OAuthOpenID credential issuance uses OAuth to deliver credentials to a wallet. That issuance flow does not require DIDs, and a credential may even be issued without key binding.6
05

What goes wrong?

A proof can verify while the trust decision remains wrong. The W3C model leaves the choice of trusted issuers, and the purposes for which they are trusted, outside its scope. Research prototypes using DIDs for agents have reached the same unresolved question: who designates trusted issuers.S1 §5.1,7

Repeated presentations can expose a pattern of use. Our reading of the credential standard found that reused credentials let verifiers link presentations, while revocation checks can let issuers observe use. Its privacy defences depend on the chosen technology and the issuer's cooperation; they are not a benefit every credential automatically supplies.8

For agents, possession also leaves an authority gap. The credential standard does not establish that a holder is authorized to present what it holds and calls for an accompanying authorization framework when credentials are used for authorization. We would ask the deployment to show that framework before treating a successful credential check as permission to act.1

06

What should a security leader ask?

  1. 01

    Which issuers do we accept, for which statements, and who maintains that list?

  2. 02

    Which DID methods can we resolve, and who can rotate or recover their keys?

  3. 03

    How do we establish that this presenter may act for the credential subject?

  4. 04

    What status checks happen when a credential is presented, and what happens if they fail?

  5. 05

    Can repeated presentations link a person or agent across services?

07

Sources

›Evidence: 8 claims from 5 briefings
  1. 1BI-FIND-2025-0021 · The W3C credential standard leaves out what agents need most: which issuers to trust, and who may presentThe W3C credential standard leaves out the two answers an agent deployment needs. Which issuers a verifier should trust is, in its words, out of scope for this recommendation, and trust is direct, never inherited along a chain. It also has no mechanism to establish that a holder is authorized to present what it holds, and says using credentials for authorization needs an accompanying authorization framework. Its only examples of a holder who is not the subject are a parent and a pet owner; software holders and presenting on someone else's behalf do not appear. The DID-based agent schemes built on it, KYA-OS and the TU Berlin prototype, both stop at the same issuer-trust question.
  2. 2BI-FIND-2026-0640 · DIF's agent spec names every AI agent by a DID; the IETF's Agent Name Service switched to domain namesKYA-OS names AI agents by decentralized identifiers and nothing else. Implementations must support did:key and did:web; no DNS name, OAuth client id or other identifier counts as an agent's identity, and the agent generates its own key, which no registry or verifier may ever hold. That is narrower than DIF described it in April 2026, when the first conformance level was to use existing OIDC and JWT identifiers; the v1.0 conformance document's Level 1 begins by deriving a did:key. It puts the Decentralized Identity Foundation's agent spec on the opposite side from the IETF's Agent Name Service, which moved from DIDs to domain names in its April 2026 revision.
  3. 3BI-FIND-2026-0641 · DIF's agent spec names every AI agent by a DID; the IETF's Agent Name Service switched to domain namesKYA-OS writes narrowing into the verifier's duties, which the Delegate SD-JWT draft does not. Each delegation is a W3C credential whose issuer must be the previous credential's subject; scopes must be equal to or narrower than the parent's at every link, the chain must end at an accountable root it calls the Responsible Party, and the verifier must check the whole chain rather than the last credential. The newer Entity Card profile fails closed on any hop that does not attenuate, with action sets as subsets and amount and time caveats that may only tighten. The rule binds only KYA-OS's own scope language, and revocation, which cascades from any ancestor, is required only at the top conformance level.
  4. 4BI-FIND-2026-0642 · DIF's agent spec names every AI agent by a DID; the IETF's Agent Name Service switched to domain namesThe specification proves which DID signed and what it was delegated, and leaves open whose DIDs to accept. It defines delegation, credential and trust registries without naming anyone to run them, makes a trust registry's rejection advisory, and puts federation between directories out of scope. A TU Berlin prototype of agents with DIDs stopped at the same point, calling the choice of trusted issuers unresolved, so for DID-named agents the open question is who runs the list of agents worth believing.
  5. 5BI-FIND-2026-0636 · SD-JWT VC is one step from an RFC, and it gives verifiers two ways to find an issuer's key, neither of them a DIDSD-JWT VC is one step from approval and names exactly two ways for a verifier to find the issuer's key: a metadata document at a well-known path under the issuer's HTTPS identifier, or an X.509 certificate chain in the credential's header. The OAuth working group has submitted revision 19 to the IESG, where it waits on area director Deb Cooley's go-ahead with the security directorate's Last Call review marked 'has issues'. Decentralized identifiers appear nowhere in the body; an early revision's change log says a DID mechanism is left to profiles and extensions, and a later one adds that the format is not the W3C Verifiable Credentials Data Model. The IETF's credential format, in other words, anchors trust in the web's names and certificates.
  6. 6BI-FIND-2025-0045 · OpenID's credential issuance standard is OAuth for wallets, and assumes a trusted wallet is an app on a phoneOpenID4VCI puts credential issuance on ordinary OAuth: the wallet is an OAuth client, the credential endpoint is a protected resource, and the hardening is the kit the desk has followed for agents, with pushed authorization requests and PKCE recommended, DPoP recommended for native wallets, and any access token living longer than five minutes required to be sender-constrained. The issuer is always an HTTPS URL. A credential's key may be named by a DID, and issuer metadata may list did methods as binding options, but nothing requires one; binding is by key, and an issuer may even issue unbound credentials. An organization that already runs an OAuth authorization server is most of the way to issuing credentials.
  7. 7BI-FIND-2025-0061 · Left to run a decentralized-ID check, AI agents finished as few as 4 runs in 100; one pair agreed to skip halfThe design proves which agent holds a key and leaves open whom to believe. Cross-domain trust rests on two assumptions the authors state: both sides trust the ledger, and each trusts the other domain's credential issuer. How trustworthy issuers are identified and designated, they write, remains unresolved; they suggest something like the EU's eIDAS trusted lists adapted for agents. A 2024 paper on IDs for AI systems stopped at the same step, so two research groups building agent identity from opposite ends have both left the trust decision to a list nobody yet runs.
  8. 8BI-FIND-2025-0023 · The W3C credential standard leaves out what agents need most: which issuers to trust, and who may presentAn agent that presents the same credential again and again is the correlation case the standard warns about: repeated presentation lets one verifier, or several colluding, link every use to one holder, and revocation lookups let the issuer watch too. The defences it names, unlinkable BBS signatures, single-use credentials and Oblivious HTTP, are advice rather than requirements and depend on the issuer's cooperation. OpenID4VCI's batch issuance of same-content credentials with different keys is the one deployed mechanism in our corpus that does this.

›In our briefings

The claim each of the newest briefings on DID / Verifiable Credentials would have you leave with.

  1. The DID standard does not require a DID to be decentralized. Version 1.1 gives, as an example, DIDs based on identifiers registered in "federated or centralized identity management systems", lists "databases of any kind" among its examples of a registry a DID can be recorded on, and does not require or suggest any specific type of registry. A method specification only SHOULD be registered in the DID Extensions list, and the spec notes that no central authority mandates which specification goes with a given method name. Mapping human-friendly names, a domain name among them, to DIDs is out of scope. We read this as making decentralization, like revocation and recovery, a property of each DID method.

    The DID standard does not require a DID to be decentralized, and leaves revocation and recovery to each methodBI-FIND-2026-0649

  2. Verifiable Intent chains three layers of signed SD-JWT credentials, tokens whose fields can be shown selectively. The issuer, a bank or payment network, binds the user's key for about a year (L1). The user signs mandates for the agent (L2): final values for an immediate purchase or, in autonomous mode, constraints plus the agent's key for a recommended 24 hours to 30 days. In autonomous mode the agent then signs a payment mandate for the network and a checkout mandate for the merchant, each valid about five minutes (L3a, L3b). The specification says "Layer 3 credentials MUST NOT contain a `cnf` claim in the payload (terminal delegation)", so the agent's credential names no next key and the agent cannot pass its authority to another agent. A verifier checks the agent's final values against the user's constraints, but only the ones disclosed to it, and it skips the allowlist check when no allowlist was disclosed. Eight constraint types are defined (merchants, line items, payees, amount range, budget, recurrence, agent recurrence, reference). Budget and the two recurrence limits need the payment network to keep a running count.

    Verifiable Intent stops an AI agent from passing its buying authority on, and v0.1 has no protocol to revoke itBI-FIND-2026-0477

  3. Vajragowni's paper puts fixed numbers on governing non-human identities (NHIs). His lifecycle algorithm treats a credential older than 90 days as stale, rotating API keys and service-account passwords and renewing certificates. It calls an identity unused for more than 180 days dormant and quarantines any NHI with no owner at once, with a ticket to assign an owner or decommission it within 14 days. Identities both orphaned and unused for more than 365 days are decommissioned. His KPI table recommends targets such as under 2% of NHIs with no assigned owner or application, over 98% of credentials rotated within the policy window, and under 4 hours on average from a decommission trigger to revocation. The paper reports no deployment or test of these values and gives no reason for choosing them. Its own dormancy KPI counts identities with no activity in 90 days, where the algorithm waits 180.

    A paper's 90- and 180-day rules for machine identities come with no test behind themBI-FIND-2026-0030

›Cited in 94 briefings

Every briefing that discusses this term, newest first.

  1. The DID standard does not require a DID to be decentralized, and leaves revocation and recovery to each method

    Article · · W3C · W3C Decentralized Identifier Working Group · Mar 5, 2026

  2. Verifiable Intent stops an AI agent from passing its buying authority on, and v0.1 has no protocol to revoke it

    Article · · Verifiable Intent (GitHub, agent-intent/verifiable-intent) · Verifiable Intent Working Group · Apr 18, 2026

  3. A paper's 90- and 180-day rules for machine identities come with no test behind them

    Article · · International Journal For Multidisciplinary Research (Vol 8, Issue 3) · Thrinaath Vajragowni · May 23, 2026

  4. OAuth's token revocation list nears RFC, and a shared list hides which token is checked

    Draft · · IETF Datatracker · Tobias Looker et al. · Aug 13, 2026

  5. EMVCo's agent card-payment draft binds no one yet and points to a spec FIDO maintains for proof of consent

    Article · · EMVCo · Aug 27, 2026

  6. W3C's JOSE and COSE rules for credentials admit there is no standard way yet to fetch a JWK from a DID URL

    Article · · W3C · Michael Jones et al. · Oct 13, 2026

  7. In SD-JWT VC, the X.509 key route does not authenticate the iss claim

    Draft · · IETF Datatracker · Oliver Terbu et al. · Aug 31, 2026

  8. A digital ID is worth little until businesses can accept it, five voices say, and four point to bank onboarding

    Field signals · · 14 voices from the field · Jun 2024 to Oct 2026

  9. Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claim

    Draft · · IETF Datatracker · Kieran Sweeney · Oct 13, 2026

  10. Agent URI draft -04 ties an agent's name to a key document and bars OAuth servers from trusting the name alone

    Draft · · IETF Datatracker · Yaswanth Narvaneni et al. · Oct 20, 2026

  11. Drake's revised draft: an agent identity outlives its operator but is never reissued once its last key is gone

    Draft · · IETF Datatracker · Christopher Drake, 1id.com · Sep 25, 2026

  12. NovaFabric seals AI agent runs as signed evidence, and its author says a signature proves a key, not a person

    Paper · · arXiv · Mohsen Seyedkazemi Ardebili · Oct 25, 2026

  13. One global registry for AI agent identities, and no one allowed to switch an identity off

    Draft · · IETF Datatracker · Christopher Drake, 1id.com · Oct 27, 2026

  14. A blockchain can't show who is behind a wallet, says Cyberport, as Hong Kong's digital-asset pilots go institutional

    Podcast · · Trust Talks: Exploring Organizational Identity in the Digital Age · Inside a Regulated Digital Asset Hub: What It Takes for Institutions to Trust Web3 · Oct 23, 2026

  15. Delegation receipts for AI agents, revision 15: holding a copy of a receipt no longer lets you extend the chain

    Draft · · IETF Datatracker · Ryan Nelson, Authproof · Oct 22, 2026

  16. ChainIT's design has the payment rail rebuild each payment's hash before money moves; the paper reports no test

    Paper · · arXiv · Jeremy Blackburn et al. · Oct 16, 2026

  17. Orange researchers would drop certificates from agent handshakes by putting post-quantum keys on a ledger

    Paper · · arXiv · Karol Budzyn et al. · Oct 16, 2026

  18. An 'approved' flag can record four different human acts, and only a check says anything about the thing approved

    Draft · · IETF Datatracker · Andreas Ehstand · Sep 27, 2026

  19. UCAN 1.0 drops the authorization server, and its editors admit a delegator cannot see later hand-offs

    Article · · GitHub (ucan-wg) · Brooklyn Zelenka et al. · Oct 2, 2026

  20. An OAuth agent profile has the token server block any widening of scope, and leaves revoking the chain optional

    Draft · · IETF Datatracker · Sanjeev Kumar, Orchestrum Technologies LLP · Aug 30, 2026

  21. A transparency-log profile for agents exposes edited and dropped records; an action nobody logged leaves no gap

    Draft · · IETF Datatracker · Pinar Emirdag et al. · Oct 12, 2026

  22. Human Delegation Provenance is now an audit record only, and one that cannot prove it is complete

    Draft · · IETF Datatracker · Asiri Dalugoda, Helixar · Oct 6, 2026

  23. A decentralized ID design for agents proves who is liable but can collect only from collateral locked in advance

    Paper · · arXiv · Ahammod Munim Ramil et al. · Oct 12, 2026

  24. OpenA2A's protocol proves an agent holds a key and scores its behaviour, but never says who it works for

    Draft · · IETF Datatracker · Abdel Fane, OpenA2A · Oct 6, 2026

  25. GLEIF's credentials vouch for a company and its officers; an AI agent's would carry only the company's assurance

    Podcast · · The Future of Identity · Alexandre Kech – Strengthening Organizational Trust with Verifiable Legal Entity Identifiers · Aug 5, 2026

  26. DIF's agent spec names every AI agent by a DID; the IETF's Agent Name Service switched to domain names

    Article · · Decentralized Identity Foundation (KYA-OS) · Dylan Hobbs et al. · Jul 29, 2026

  27. The first DID methods DIF and Trust over IP recommend are both web-based; blockchain ones are still in review

    Article · · Decentralized Identity Foundation · Decentralized Identity Foundation et al. · Jun 19, 2026

  28. In Accenture's agent-delegation design, no outside party checks the user, and revocation is left open

    Paper · · arXiv · Srikumar K. Subramanian et al. · Sep 25, 2026

  29. A survey rates only workload identity ready for agents, and argues no standard can tell what an agent intends

    Paper · · arXiv · Takumi Otsuka et al. · Apr 25, 2026

  30. Stripe and Tempo propose paying for an HTTP request the way you log in to one, and leave the agent out

    Draft · · IETF Datatracker · Brendan Ryan et al. · Sep 9, 2026

  31. An integrator builds agent access on OpenID Connect and OAuth; clients want one view across many MCP gateways

    Podcast · · Identity Unfiltered · CIAM, Agentic AI & the Future of Identity · Oct 5, 2026

  32. McDonald's protects 2.2 million mostly franchise workers with a printed sheet of codes

    Podcast · · Identity Decoded | The Identity Security Podcast · Identity Security on the Frontlines: How McDonald’s Built a Scalable Strategy · Sep 23, 2026

  33. A county CTO says identity is already critical infrastructure, and nobody will own it until someone declares it

    Podcast · · Identity at the Center · #452 - When Identity Becomes Critical Infrastructure with Shea McGrew · Oct 5, 2026

  34. OAuth handles an agent in one company, working while you watch. The OpenID paper says that is where it stops.

    Article · · OpenID Foundation · Tobin South et al. · Oct 2025

  35. The Agent Name Service draft was rebuilt on DNS, with a GoDaddy author, and no standards body has adopted it

    Draft · · IETF Datatracker · Scott Courtney et al. · Apr 13, 2026

  36. xChk ships human OIDC portfolios and a separate agent approval API

    Paper · · arXiv · Sean MacGuire · Jul 15, 2026

  37. Glide pitches the SIM card's key as the replacement for SMS codes, resting its AAL3 claim on hardware alone

    Podcast · · The Future of Identity · Eran Haggiag – Rebuilding Digital Identity Around SIM Cards for the AI Era · Jun 23, 2026

  38. Huawei researchers add agent delegation to an existing permission model and prove each action traces to a human

    Paper · · arXiv · Amjad Ibrahim et al. · Jun 2, 2026

  39. An AI agent's refused action can be approved only by someone its mandate names, says a revised draft

    Draft · · IETF Datatracker · Lars Kersten Kroehl, CryptoKRI GmbH · May 21, 2026

  40. A preprint cuts off every sub-agent within about 40 seconds of its parent stopping, with no revocation server

    Paper · · arXiv · Saurabh Deochake, SentinelOne · May 20, 2026

  41. A proposed standard for agent authority advises against delegating it across company lines by default

    Paper · · arXiv · Partha Madhira · May 12, 2026

  42. AgentDID checks an AI agent's identity and readiness in 13.5 seconds, but was tested on at most 50 agent pairs

    Paper · · arXiv · Minghui Xu et al. · Apr 28, 2026

  43. A 'decentralized' agent identity draft blocks risky calls when its Registry is down, and did:aip is taken

    Draft · · IETF Datatracker · Paras Singla · Apr 17, 2026

  44. A proposed consent log for A2A and MCP records the terms an agent accepted, but the agent fills it in itself

    Paper · · arXiv · Ravi Kiran Kadaboina · Apr 16, 2026

  45. Most agent-payment protocols leave the agent's own reasoning out of scope, a survey of 142 works finds

    Paper · · arXiv · Qian'ang Mao et al. · Apr 15, 2026

  46. A proposed record of who authorized an AI agent's actions cannot yet prove which agent took them

    Paper · · arXiv · Asiri Dalugoda, Helixar · Apr 6, 2026

  47. AIP puts an agent's delegation chain in the token; its perfect score against attacks tests rules only AIP has

    Paper · · arXiv · Sunil Prakash · Mar 25, 2026

  48. A cross-domain agent trust draft requires the call chain, then leaves its format to a version that never came

    Draft · · IETF Datatracker · Daniel King et al. · Mar 15, 2026

  49. A DIF report says login-era identity systems struggle when AI agents work across companies, and lists five needs

    Article · · Decentralized Identity Foundation · Damian Glover · Mar 12, 2026

  50. An Alibaba and Cisco draft signs what the user was shown, then enforces a policy the agent wrote

    Draft · · IETF Datatracker · Dapeng Liu et al. · Feb 28, 2026

  51. Google DeepMind wants agent authority to narrow at every hand-off, and misreads MCP's scopes on the way

    Paper · · arXiv · Nenad Tomašev et al. · Feb 12, 2026

  52. A Cisco-led agent draft puts the verifiable credential at a public URL, with selective disclosure a SHOULD it does not yet show

    Draft · · IETF Datatracker · Jean Diaconu et al. · Feb 6, 2026

  53. A secure-MCP proposal registers every agent and tool centrally, and never mentions MCP's own OAuth rules

    Paper · · arXiv · Xinyi Hou et al. · Feb 1, 2026

  54. A proposed agent:// name survives a move between clouds but not a change of job

    Paper · · arXiv · Roland R. Rodriguez et al. · Jan 21, 2026

  55. A Shenzhen paper ties each AI agent to a verified owner and its exact code, at 15 to 40 seconds of proving a step

    Paper · · arXiv · Zibin Lin et al. · Dec 19, 2025

  56. FIDO moves past passkeys to certify identity wallets, so issuers can check a wallet before provisioning to it

    Article · · FIDO Alliance · FIDO Alliance · Dec 4, 2025

  57. CORAL's public agent server rejected forged messages but let a wrong key open the session stream

    Paper · · arXiv · Yedidel Louck et al. · Nov 5, 2025

  58. Credential wallet draft makes holder binding a MUST; its own SD-JWT profile names no way to provide it

    Draft · · IETF Datatracker · Leif Johansson et al. · Nov 4, 2025

  59. AAGATE maps NIST's AI risk framework onto Kubernetes, and its paper never times the 'millisecond kill switch'

    Paper · · arXiv · Ken Huang et al. · Oct 29, 2025

  60. SPICE includes software agents as credential issuers, holders and verifiers

    Draft · · IETF Datatracker · Brent Zundel et al. · Oct 20, 2025

  61. Left to run a decentralized-ID check, AI agents finished as few as 4 runs in 100; one pair agreed to skip half

    Paper · · arXiv · Sandro Rodriguez Garzon et al. · Oct 1, 2025

  62. FIDO pairs passkeys with digital ID credentials but spells out steps only for moderate assurance

    Article · · FIDO Alliance · Christine Owen et al. · Sep 22, 2025

  63. Agentic JWT binds an agent's token to its task and its own code, but it is one author's draft, not a standard

    Paper · · arXiv (cs.CR preprint 2509.13597) · Abhishek Goswami · Sep 16, 2025

  64. OpenID's credential issuance standard is OAuth for wallets, and assumes a trusted wallet is an app on a phone

    Article · · OpenID Foundation · Torsten Lodderstedt et al. · Sep 16, 2025

  65. A DeepMind paper puts agent markets on DIDs and credentials but leaves delegation unspecified

    Paper · · arXiv · Nenad Tomašev et al. · Sep 12, 2025

  66. A blockchain-agent survey's blueprint has no agent identity; risk scores and a human signature guard the wallet

    Paper · · arXiv · Nicolò Romandini et al. · Sep 8, 2025

  67. The Aegis agent-security paper's 2.79-second proof time is a benchmark-based target its simulation reports back

    Paper · · arXiv · Sai Teja Reddy Adapala et al. · Aug 22, 2025

  68. A 'provable' bound on agent-memory injection assumes independent layers and fills in none of its detection rates

    Paper · · arXiv · Ken Huang et al. · Aug 17, 2025

  69. ERC-8004 stopped requiring agents to approve who rates them in January and admits fakes can inflate ratings

    Article · · Ethereum Improvement Proposals · Marco De Rossi et al. · Aug 13, 2025

  70. DIRF credits its anti-cloning controls with a jump from near 0% to over 90%; no test in the paper applies them

    Paper · · arXiv · Hammad Atta et al. · Aug 2, 2025

  71. A Tsinghua blockchain design for AI agents backs its defense claims with a table and reports no attack run

    Paper · · arXiv · Zhenhua Zou et al. · Aug 2, 2025

  72. OpenID4VCI's third interop test: 87% of tested pairs passed, on counts the post does not reconcile

    Article · · OpenID Foundation · Gail Hodges · Jul 24, 2025

  73. OpenID4VP 1.0 makes verifiers re-check what a wallet returns, and cannot show a holder acting for someone else

    Article · · OpenID Foundation · Oliver Terbu et al. · Jul 9, 2025

  74. China Mobile's agent-identity draft had one MUST, binding each agent to its user, and no mechanism to do it

    Draft · · IETF Datatracker · Kehan Yao et al. · Jul 1, 2025

  75. AI governance researchers argue OAuth tokens can't link an agent's work across services or sub-agents

    Paper · · arXiv · Alan Chan et al. · Jun 19, 2025

  76. ETDI would sign MCP tool definitions against lookalikes, but leaves open who vouches for the publisher's key

    Paper · · arXiv · Manish Bhatt et al. · Jun 2, 2025

  77. A Cloud Security Alliance paper wants AI agents on a new identity stack, with OAuth kept for simple cases

    Paper · · arXiv (cs.CR preprint 2505.19301) / Cloud Security Alliance · Ken Huang et al. · May 25, 2025

  78. W3C Data Integrity proofs name their purpose, so a verifier can refuse a credential signature offered as a login

    Article · · W3C · Manu Sporny et al. · May 15, 2025

  79. W3C's Controlled Identifiers 1.0 lets an HTTPS URL publish its keys, and whoever can edit that document is its controller

    Article · · W3C · Manu Sporny et al. · May 15, 2025

  80. The first Agent Name Service paper puts capabilities in the name and leaves protocol translation and governance out

    Paper · · arXiv (cs.CR preprint 2505.10609) / OWASP Agentic Security Initiative · Ken Huang et al. · May 15, 2025

  81. The W3C credential standard leaves out what agents need most: which issuers to trust, and who may present

    Article · · W3C · Manu Sporny et al. · May 15, 2025

  82. Decentralized AI agents have no off switch by default; builders say cutting API access still stops most of them

    Paper · · arXiv · Botao Amber Hu et al. · May 14, 2025

  83. Germany's wallet builders say the state certifies and lists wallets, and users pick any listed one, Google's too if it meets the rules

    Podcast · · Identity at the Center · #354 - Kristina Yasuda & Torsten Lodderstedt on the EUDI Wallet and its Global Impact · First aired May 7, 2025 · Jun 9, 2025

  84. A 'zero trust' registry against AI tool squatting still has to trust its administrator

    Paper · · arXiv · Vineeth Sai Narajala et al. · Apr 28, 2025

  85. An A2A threat model flags the Agent Card as an attack path, and its example flow never passes on the user's identity

    Paper · · arXiv · Idan Habler et al. · Apr 23, 2025

  86. LOKA lets AI agents own their identities and vote on ethics, without naming anyone they act for

    Paper · · arXiv · Rajesh Ranjan et al. · Apr 15, 2025

  87. An MIT-led paper has the user sign the agent's delegation token and lets OpenID Connect vouch for both parties

    Paper · · arXiv · Tobin South et al. · Jan 16, 2025

  88. Lockstep’s 2023 wallet proposal puts credential acceptance outside the wallet

    Video · · Identiverse - A CRA Resource · Uploaded June 17, 2024; talk recorded May 31, 2023

  89. The 2024 'IDs for AI' paper puts the ID on a chat session, not a model, and leaves trust in its issuer to others

    Paper · · arXiv · Alan Chan et al. · Jun 17, 2024

  90. DIF's Presentation Exchange lets a verifier ask for credentials, but not ask who is presenting them for whom

    Article · · Decentralized Identity Foundation · Daniel Buchner et al. · Apr 25, 2024

  91. JWT sub and sub_id must identify the same subject

    Article · · RFC Editor · Annabelle Backman et al. · Dec 6, 2023

  92. In 2025, Kuppinger said Microsoft was underestimated and expected EU wallets to make verification reusable

    Podcast · · Identity at the Center · #326 - IAM in 2025 with Martin Kuppinger · Jan 20, 2025

  93. Microsoft's 2023 keynote: workload standards should carry identity across service calls first, policy later

    Video · · Identiverse - A CRA Resource · Recorded May 31, 2023 · Uploaded June 17, 2024

  94. A 32-author 2024 paper asks for a credential that proves you are human and nothing more

    Paper · · arXiv · Steven Adler et al. · Aug 15, 2024

In our columns

  1. A signed AI agent name is an IP address with a key, not a login

    Column · Oct 30, 2026

  2. However an AI agent is named, the trust behind it runs through a web domain

    Column · Oct 19, 2026

  3. Gartner's identity curve needs a closer look at what has actually shipped

    Column · Oct 15, 2026

  4. Machine identities lack a shared schema, so buy discovery tools that export their data

    Column · Sep 17, 2026

Related terms

What appears alongside DID / Verifiable Credentials in the same briefings.

did:keyKYA-OSDecentralized Identity FoundationA2AGDPRSD-JWTJWTModel Context Protocol

Was this explainer useful?

Share this explainer

Pass it on. Every claim has a permanent number, so what you share today still points at the same words next year.