Skip to content
IAM Insights

Explainers · Glossary

SPIFFE

SPIFFE/SPIRE

By Boston Identity · Explained · 7 min read

The short answer

SPIFFE is an open standard that gives every running service a verifiable identity. A local API hands each workload a short-lived certificate that names it, so services can prove who they are to each other without shared secrets.

What it is
A standard
Defined by
SPIFFE project (CNCF), stable specifications
Who it's for
Teams running services or AI agents that call each other
In our briefings
101By month of the source: 2025-12: 1, 2026-01: 1, 2026-02: 5, 2026-03: 6, 2026-04: 8, 2026-05: 8, 2026-06: 8, 2026-07: 14, 2026-08: 3, 2026-09: 17, 2026-10: 18, 2026-11: 11by month, last 12
01

What is SPIFFE?

Services call each other all day, and each needs to know who is calling. The usual answer has been a long-lived shared credential, such as an API key or a service-account password, and CyberArk now issues SPIFFE identity documents in its place.1 SPIFFE, the Secure Production Identity Framework For Everyone, is a framework for bootstrapping and issuing identity to services across different environments and organizations.S1 Abstract Its certificate specification says that perhaps its most important job is securing communication from one process to another.S2 §1

Every workload gets a SPIFFE ID, a URI such as spiffe://staging.example.com/payments/web-fe.S1 §2.2 The part after spiffe:// is the trust domain, which names the system that issued the identity.S1 §2 The path names the workload, and the administrator decides what it means.S1 §2.2 A workload proves its ID with a SPIFFE Verifiable Identity Document (SVID), signed by an authority inside its trust domain.S1 §3 An SVID reuses an existing format, an X.509 certificate, a JSON Web Token (JWT) or a workload identity token (WIT), and each format has its own specification.S1 §3.3

Not the same as SPIRE. SPIFFE is the set of specifications. SPIRE is software that runs them, such as the agent on each host that Uber's services ask for their credentials.2

02

How does SPIFFE work?

Take the specification's own example: a web front end and a MySQL database, both part of a payments service in the staging.example.com trust domain.S1 §2.2 The front end wants to call the database, and each service can reach a local Workload API.

Figure steps
  1. The front end asks the Workload API
  2. The API hands back an identity
  3. The two services open mutual TLS
  4. Each side checks the other's SVID
  5. A partner's bundle lets calls cross domains
  6. The API streams rotations and revocations
Fig. 1

Neither service starts with a secret: each gets its certificate and key from the local Workload API, and trusts the other only through the bundle for the other's trust domain.

Drawn from the SPIFFE ID, X509-SVID, Workload API and bundle specifications

Swipe for the full figure →

  1. The front end asks the Workload API

    The front end connects to the local Workload API and calls FetchX509SVID.S3 §5.1 It sends no password or key. The endpoint does no direct client authentication. Its implementation identifies the caller through out-of-band checks and decides what to serve.S3 §4.1 A caller that is entitled to nothing should get a PermissionDenied error.S3 §5.2.1

  2. The API hands back an identity

    The reply carries one X509-SVID for each identity the caller is entitled to.S3 §5.2.1 Each comes with its private key and the trust bundle for its trust domain.S3 §5.1 A bundle is the set of keys used to check that domain's SVIDs.S4 §3 The certificate holds the SPIFFE ID in exactly one URI subject alternative name (SAN), and a validator must reject an SVID with more than one.S2 §2 The database gets its own SVID the same way, since the API serves any process that can reach it.S3 §4.1

  3. The two services open mutual TLS

    The front end connects to the database over TLS, and each side presents its X509-SVID as its certificate.S2 §1 The database checks the client, and the front end checks the server.S3 §4.6 A workload certificate should list its extended key usages, and if it does, it must allow both server and client use.S2 §4.4

  4. Each side checks the other's SVID

    Each side picks the bundle for the trust domain in the peer's SPIFFE ID and validates against it.S3 §4.6 If it holds no bundle for that domain, the peer is untrusted. The chain check must follow standard X.509 path validation (RFC 5280).S2 §5.1 Then come the SPIFFE checks: the certificate must not be a signing certificate, the scheme must be spiffe, and the ID must have a path.S2 §5.2 A signing certificate may issue further signing certificates in other trust domains,S2 §3.2 and the name constraints that would stop it are optional.S2 §4.2 Choosing the bundle by the peer's trust domain is what stops one domain's authority from vouching for another domain's IDs, which is why the Workload API specification requires that choice.S3 §4.6 If one root key is shared across trust domains, the database's authorization rules also need to check the trust domain explicitly.S4 §6.2

  5. A partner's bundle lets calls cross domains

    To trust a service in another trust domain, a workload needs that domain's bundle. The API can return bundles for foreign trust domains, keyed by the domain's ID.S3 §5.1 The specification calls these federated bundles the main way trust domains connect.S3 §4.6 A bundle is a standard JSON Web Key (JWK) set, and each key is marked with the kind of SVID it checks.S4 §4,S4 §4.2.2 Bundles are meant to pass between SPIFFE control planes, and the Workload API carries their updates to workloads.S4 §3

  6. The API streams rotations and revocations

    The workload should keep its connection open.S3 §4.2 The API sends a new message when something changes, such as a rotated SVID or an updated revocation list, and every message must carry the full set of information.S3 §4.3 Anything left out counts as withdrawn: a missing bundle should be unloaded, and a dropped SVID should no longer be used.S3 §4.4,S3 §5.2.1 Root keys rotate through the bundle too. In the specification's example, a replacement certificate is added well before the old one expires.S4 Appendix A

03

Why does SPIFFE matter now?

AI agents have reopened the question SPIFFE answered for services. Six engineers from six competing companies wrote an IETF draft arguing that an agent is a workload, and that WIMSE, SPIFFE, OAuth and the OpenID Shared Signals Framework already cover it. The IETF's workload identity working group (WIMSE) adopted that work on 15 September 2026.3 The draft says an agent must hold short-lived, key-bound credentials, lists a SPIFFE SVID as one allowed form, and calls static API keys an antipattern.4 An April 2026 survey graded WIMSE with SPIFFE as the only agent-identity work ready for production, though the grades are the three authors' own judgement.5

The standards bodies are building on SPIFFE. The IETF's workload certificate is an ordinary X.509 certificate with one URI SAN, which the draft calls fully compatible with the X509-SVID.6 An OAuth working-group draft lets a workload sign in to an authorization server with its SPIFFE credential instead of a client secret, and one implementation, Keycloak in preview, supports part of it.7 NIST's National Cybersecurity Center of Excellence lists SPIFFE among the existing standards it is considering for agent identity.8

Uber ran SPIFFE and SPIRE under 4,500 services on hundreds of thousands of hosts by 2023.9 When its agent platform went live, it still could not trace an automated pull request back to the engineer who started the chain.10 We read the two Uber accounts together as evidence that workload identity is necessary for agents and not sufficient.9 CyberArk's Secure Workload Access became generally available on 28 May 2026, issuing SPIFFE identity documents to Kubernetes and Unix workloads.1

04

Where does it fit?

Compared withHow SPIFFE differs
WIMSE Workload IdentifierThe IETF's workload identifier is a looser URI syntax that existing SPIFFE IDs pass unchanged, and it registers its own wimse scheme. Unlike SPIFFE, it does not say how identifiers are issued or verified.11
OAuth SPIFFE Client AuthenticationThis OAuth draft builds on SPIFFE: it lets a workload present its SPIFFE credential to an authorization server in place of a client secret.7 The four SPIFFE documents we read do not mention OAuth.
JWTA JWT is a general token format. SPIFFE uses it as one SVID type, the JWT-SVID, which a workload requests for a named audience and a validator checks against the bundle for the subject's trust domain.S3 §6.2.1,S3 §6.3
OAuthIn WorkOS's account, a standard like SPIFFE issues the credential after the workload proves what it is, and runtime authorization is left to OAuth.12
05

What goes wrong?

  • Trust domain names are self-registered. Two operators can pick the same name, and colliding domains cannot federate, so the specification advises a registered domain name as a suffix.S1 §2.1,S1 §2.1.1
  • An SVID's contents count as true for its whole lifetime. Revoking a claim inside one means waiting for every copy to expire, so the specification advises leaving out things that change, such as roles or group membership.S1 §4.1.1
  • A partner domain's word is only as good as that domain. A compromised domain could claim to have authenticated any user it wishes, and a role called admin in one domain means nothing in another.S1 §4.1.4,S1 §4.1.2
  • A root key shared across trust domains blurs them: unless authentication checks the domain name, an SVID from any of them can pass. The specification strongly advises one key per domain.S4 §2,S4 §6.2
  • Slow bundle refresh slows revocation. Without a refresh hint, clients should poll often, around every five minutes in the specification's example.S4 §6.1
  • The specifications leave how the endpoint recognizes a caller to the implementation.S3 §4.1 At Uber, that check is mainly a label from a trusted launcher, and the post describes no measurement of the running code.2
  • The OAuth draft that joins SPIFFE to OAuth does not discuss a compromised trust domain or SPIRE server.13 The OpenID Foundation's paper argues that SPIFFE's identity depends on controlling the infrastructure, so it does not cross into another organization.14
06

What should a security leader ask?

  1. 01

    Which of our services still prove who they are to each other with a stored key or password?

  2. 02

    How does our Workload API decide which process is calling, and who can change that registration data?

  3. 03

    Which partner trust domains do our workloads accept, and how fast does a revoked key drop out of their bundles?

  4. 04

    Do our authorization rules check the trust domain in a SPIFFE ID, or only the path?

  5. 05

    For our AI agents, what records the person who started each task, since an SVID names only the workload?

07

Sources

›Evidence: 14 claims from 11 briefings
  1. 1BI-FIND-2026-1186 · Three vendors put MCP controls in three different places in May, and only Auth0's was GAMay's machine-identity releases split between replacing stored credentials and managing them. CyberArk's Secure Workload Access went GA on May 28, issuing SPIFFE identity documents to Kubernetes and Unix workloads in place of long-lived shared credentials. Microsoft's provisioning service can now reach SAP SuccessFactors with workload identity and short-lived tokens instead of a stored password, in public preview. Other releases still manage stored secrets: AWS Secrets Manager now rotates Datadog keys and Snowflake tokens, Okta's service account management for Privileged Access went GA, and CyberArk's Kubernetes discovery agent now collects secret values by default.
  2. 2BI-FIND-2023-0007 · Uber had workload identity on a quarter-million nodes two years before agents; certificates were the easy partWhat a workload is, in Uber's system, is whatever the launcher said it was. Identity reaches developers as an Auth library in Go and Java that subscribes to the local SPIRE agent, attaches the credential in RPC middleware and verifies it on the way in; Uber chose a library over a sidecar or binary because its monorepos and shared RPC frameworks made the library the cheaper path to every service. Attestation is 'primarily' container-based: launchers, themselves attested, apply labels naming the service, environment and platform, and the agent maps those labels to a registered entry. No measurement of the running code is described. The IETF's credentials draft leaves attestation out of scope, and this is what out of scope looks like at a large operator: a trusted launcher's label.
  3. 3BI-FIND-2026-0178 · Six rival vendors wrote the agent-identity draft the IETF just adopted, and it invents nothingSix engineers from six companies that compete with each other, Defakto, AWS, Zscaler, Ping Identity, OpenAI and Okta, wrote one draft whose first claim is that everyone else is 'reinventing existing mechanisms unaware of applicable prior art', and whose answer is to define no new protocol at all: an agent is a workload, and WIMSE, SPIFFE, OAuth 2.0 and the OpenID Shared Signals Framework already cover it. On 15 September 2026 the WIMSE working group adopted the work as draft-ietf-wimse-aims, 'AI Identity Management System', with the same six authors; the datatracker now marks the revision we read as replaced. That is the bet Highflame's ZeroID makes with its eighteen implemented specifications, the one Uber made by tracking this draft, and the one the summit's practitioner recommended from the stage: the agent-identity layer is assembled, not invented, and the assembly now has a working group.
  4. 4BI-FIND-2026-0179 · Six rival vendors wrote the agent-identity draft the IETF just adopted, and it invents nothingThe rule Sierra built its cloud agent around is now normative text. An agent MUST have one WIMSE identifier and short-lived credentials bound to a key, a Workload Identity Token, an X.509 workload certificate or a SPIFFE SVID; 'static API keys are an antipattern for agent identity'; and the language model inside the agent never holds the credential, which the draft says 'prevents the Large Language Model from using, exposing, or being manipulated via prompt injection into disclosing the credentials'. Sierra's harness has no network access and a supervisor swaps the credential in; the summit's vendors said never hand an agent a raw token; Anthropic's evaluation agent improvised its own PyPI account because nothing stopped it. Three months after the incident and two after the post, the boundary is written down by six vendors at once.
  5. 5BI-FIND-2026-0643 · A survey rates only workload identity ready for agents, and argues no standard can tell what an agent intendsIn the authors' grading of ten standards documents for agent identity, only the IETF's workload-identity work (WIMSE) with SPIFFE counts as available for production; SAML is graded incompatible, and the rest, including OAuth extensions for agents and the AI agent authentication draft whose security section still reads 'TODO', are partial or directional. The grades are the authors' own judgement over about 80 sources found by keyword search, with no scoring method or second rater, so they record what three researchers concluded in April 2026.
  6. 6BI-FIND-2026-0375 · The IETF's workload credential binds one key to one name and says nothing else about the software holding itHalf of this draft is already running and the other half is not, and the draft says so. The Workload Identity Certificate is an ordinary X.509 certificate with the workload identifier in exactly one URI SubjectAltName, which the implementation section calls 'fully compatible with the X509-SVID' that SPIFFE deployments have issued for years; SPIFFE deployments simply use their spiffe URIs as the subject. The new piece is the JWT form, and the self-reported status reads beta at SPIFFE, alpha at Defakto Security and research at Teleport, against production for the certificate at all three. The IETF is standardizing the certificate the industry already has and attaching a token nobody has shipped yet.
  7. 7BI-FIND-2026-0316 · A draft lets a workload sign in to OAuth with its SPIFFE identity, not a secret; one preview supports part of itAn OAuth working-group draft lets a workload authenticate to an authorization server with the identity its platform issued, in place of a stored secret. It covers three forms of SPIFFE credential: a signed token presented under a new assertion type, a certificate presented over mutual TLS, and a key-bound workload token with a proof of possession. A server must advertise which it accepts and support at least one. This is the missing piece between workload identity and OAuth that speakers in this corpus have been calling duct tape, now on the standards track. One implementation is listed, Keycloak, in preview, for the first of the three forms.
  8. 8BI-FIND-2026-0686 · NIST's agent identity work starts from OAuth, SPIFFE and SCIM and defers agents from outside the enterpriseNIST's National Cybersecurity Center of Excellence framed agent identity in February as a question of applying what exists, not inventing something new. Its concept paper lists OAuth, OpenID Connect, SPIFFE, SCIM, the NGAC access-control standard and MCP as candidates it is "currently considering", and names none of the agent-specific IETF work, such as ID-JAG, the WIMSE drafts or token exchange for agents. It keeps human delegation and logging in scope and leaves agents from outside the enterprise for a later effort. The next steps are a possible project description and a lab-built practice guide, with no date; the comment period closed on 2 April.
  9. 9BI-FIND-2023-0006 · Uber had workload identity on a quarter-million nodes two years before agents; certificates were the easy partUber ran standards-based workload identity at fleet scale well before it had an agent problem. The post puts SPIFFE and SPIRE under '4,500 services running on hundreds of thousands of hosts across four clouds', later 'over a quarter million nodes', with the team involved in SPIRE 'since the early stages of the project in mid-2018'. The reasons are the ones the IETF's architecture draft would later write down: perimeter trust fails for ephemeral workloads, and 'provider-specific identity solutions are a non-starter' across clouds. This is the deployment that was already in place when Uber's agent platform went live in 2025 and still found a delegation gap, which is the strongest evidence in the corpus that workload identity is necessary for agents and not sufficient.
  10. 10BI-FIND-2026-0082 · Uber built the delegation chain the rest of the field calls unsolvedThe agent identity gap is a delegation gap, and a mature zero-trust shop hits it just as hard as anyone. Uber already had workload identity on SPIFFE/SPIRE and a service mesh when its agent platform went live in early 2025, and still could not trace an automated pull request back to the on-call engineer who set the chain in motion, because each hop dropped the context of the one before. The post's observability screenshot then shows the same three-agent example as a real session trace with a user attached to every hop. The panels and papers we have covered say the same thing from the outside; this is the first account from inside a deployment.
  11. 11BI-FIND-2026-0386 · The IETF names a workload by a web address on the organization's domain; the verifier's allowlist decides trustThe IETF's workload name is a URI and almost nothing about it is fixed. It must be absolute with a non-empty authority, that authority is the trust domain and 'SHOULD be a fully qualified domain name belonging to the organization', and it 'MUST NOT contain a query component, a fragment component, user information, or a port component'. The scheme is not defined: SPIFFE's spiffe scheme fits, and the draft registers a new wimse scheme whose path 'is not interpreted by this specification'. Implementations must accept names up to 2,048 bytes. Section 1 says the document 'does not prescribe how identifiers are issued or verified', and a workload is defined as microservices, containers, virtual machines and serverless functions; AI agents are not mentioned. This is a syntax that lets existing SPIFFE names pass unchanged, with the organization's DNS name as the one load-bearing part.
  12. 12BI-FIND-2026-0408 · Three companies merged SCIM's two agent drafts, but no working group has taken it up and it expires in DecemberThe year drew a clearer line around what SCIM provisions, without settling how far it reaches. The working group's framing splits public agent identity from ephemeral workload identity, and the post leaves open whether SCIM should cover both or only the first. In WorkOS's phrasing, SCIM 'provisions the right to an identity, not the identity itself': it says which agents exist and whether they are still active, a standard like SPIFFE issues the credential after the workload proves what it is, and runtime authorization is left to OAuth 'by design, not by omission'. The SCIM working group's IETF 125 slides go further than the post: they propose that SCIM drive both OAuth client registration and SPIFFE registration entries through the same identity provider. We read this as meaning that deprovisioning an agent through SCIM ends its right to exist, and reaches its credentials only if the identity provider also wires SCIM to those registrations; running sessions keep their own expiry either way.
  13. 13BI-FIND-2026-0318 · A draft lets a workload sign in to OAuth with its SPIFFE identity, not a secret; one preview supports part of itWhat the draft leaves out is most of what a security team would ask. How an authorization server comes to trust a SPIFFE trust domain is 'out of band' and out of scope. Its own security text is one rule against accepting a token just because its issuer's keys verify it; replay protection goes no further than an expiry check and a single-audience rule, key rotation is a recommendation to poll, and a compromised trust domain or SPIRE server is not discussed. It does not mention AI agents or delegation at all. It replaces the secret. Everything about what the authenticated workload may then do is someone else's document.
  14. 14BI-FIND-2025-0057 · OAuth handles an agent in one company, working while you watch. The OpenID paper says that is where it stops.The paper finds every existing way of identifying an agent as a client incomplete. Dynamic client registration 'creates a large number of anonymous clients'. Client ID Metadata Documents fix that and are 'still insufficient as a workload identity'. SPIFFE gives a strong workload identity that depends on controlling the infrastructure, so it does not cross into another organization. Vendor systems such as Microsoft Entra Agent ID and Okta's are named as proprietary, with limited interoperability. In the year since, the Model Context Protocol has made metadata documents its preferred method and an IETF draft has joined them to SPIFFE.

›In our briefings

The claim each of the newest briefings on SPIFFE would have you leave with.

  1. Okta's and Microsoft's competing SCIM drafts for AI agents have given way to one joint draft, and no IETF working group has taken it up. In June 2026 Mark Wahl and Pamela Dingle of Microsoft, Danny Zollner of Okta and Ismael Kazzouzi of Nextident filed draft-wzdk-scim-agent-resource-00, an informational draft that defines the agent as its own SCIM resource type. WorkOS calls the joint authorship "a real signal of forming consensus". The datatracker lists it as an individual submission that replaces neither earlier draft; both of those have expired, and this one expires on 7 December 2026. WorkOS says the October 2025 Okta draft by Abbey and Cohen modelled agents through an "agentic application" instead of as their own resource. As filed, that draft already defines Agent as its own resource type, related to an agentic application as owned, authorized or guest, and the SCIM working group's IETF 125 slides say both drafts represented agents as a resource type. The joint draft's text never mentions the Okta draft or agentic applications, and it has none of the Okta draft's certificate, protocol or OAuth configuration attributes, so we would not call it a merger of the two texts. An agent governance extension from an Oracle author, draft-kushwaha-scim-agent-governance-00, filed in July 2026, was also active on the datatracker; the post does not mention it.

    Okta and Microsoft now co-author one SCIM draft for AI agents, but no working group has taken it upBI-FIND-2026-0407

  2. Uber runs per-hop token exchange with a full actor chain for AI agents in production, and has published sample tokens and a latency chart. Its security token service gives each hop a fresh token scoped to one audience, with a lifetime "in the order of minutes" (both sample payloads expire 300 seconds after issue), and the token carries the attested chain from the engineer to the agent making the call. Uber says "the system has been adopted by thousands of internal agents" and reports the exchange API's P99 as "consistently below 40 milliseconds" under current load; its chart sits mostly under 10 ms, with one spike reaching 40. The post covers this identity layer. Uber frames dynamic access control and a unified policy enforcement plane as its direction on top of it, and names session and intent claims as additions the token could carry in the future.

    Uber runs per-hop token exchange with a full actor chain for AI agents in productionBI-FIND-2026-0081

  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 101 briefings

Every briefing that discusses this term, newest first.

  1. Okta and Microsoft now co-author one SCIM draft for AI agents, but no working group has taken it up

    Article · · WorkOS Blog · Maria Paktiti · Sep 30, 2026

  2. Uber runs per-hop token exchange with a full actor chain for AI agents in production

    Article · · Uber Engineering blog · Matt Mathew et al. · May 21, 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. AAuth gives every agent its own key, then keeps the delegation chain out of the token on purpose

    Draft · · IETF Datatracker · Dick Hardt, Hellō · Sep 25, 2026

  5. Transaction Tokens carry who started a request through internal services and define no list of the hops in between

    Draft · · IETF OAuth Working Group · Atul Tulshibagwale et al. · Jul 30, 2026

  6. The JWT an agent or other OAuth client signs in with may name only the server's issuer, never the token endpoint

    Draft · · IETF Datatracker · Michael B. Jones et al. · Apr 28, 2026

  7. Palo Alto's own survey puts machine identities at 109 per person, up from 82 a year earlier

    Podcast · · Packet Protector · PP115: Palo Alto Networks: Reality of 109 to 1: Securing Machine Identities and AI Agents (Sponsored) · Jun 23, 2026

  8. OAuth token exchange does not guarantee a sub-agent stays inside its parent's access

    Article · · Control Plane by Karl McGuinness · Karl McGuinness · Mar 18, 2026

  9. SPIFFE's JWT-SVID spec says its own tokens can be replayed, and sets no hard cap on how long they live

    Article · · SPIFFE project (GitHub) · SPIFFE project · Jul 1, 2026

  10. An agent-mandate draft checks who may delegate, and now intends to carry its chain in Asor's format

    Draft · · IETF Datatracker · Claudiu Hulea, Braincap SRL · Oct 24, 2026

  11. Agent-audit draft now admits a call on the principal and the current agent, as RFC 8693 requires

    Draft · · IETF Datatracker · Shameer Thaha et al. · Oct 19, 2026

  12. Two WIMSE participants say AI agents are workloads, and existing standards already cover most of the job

    Video · · Identerati Office Hours · Nov 6, 2026

  13. Okta and Google tie an agent's calls to the user, while GitHub's shared Copilot automations now act as Copilot

    Release tracker · · 11 vendors' release notes · Oct 2026

  14. An OAuth client gets to name who vouches for it, and removing one waits on server caches

    Draft · · IETF Datatracker · Karl McGuinness · Sep 28, 2026

  15. SPIFFE trust domain names have no registrar, and the core spec warns that roles in a credential can go stale

    Article · · SPIFFE project (GitHub) · SPIFFE project · Jul 1, 2026

  16. An OAuth draft lets the API, not the agent, describe the approval, and now name the key the token is bound to

    Draft · · IETF Datatracker · Yaroslav Rosomakho et al. · Jun 25, 2026

  17. WIMSE's agent draft adds how an agent's access ends and where policy engines fit, without a new requirement

    Draft · · IETF WIMSE Working Group · Pieter Kasselman et al. · Sep 15, 2026

  18. OAuth's URL client ID draft tightens server rules in -03; the client is still whoever controls the URL

    Draft · · IETF Datatracker · Aaron Parecki et al. · Jul 6, 2026

  19. OAuth's SPIFFE draft opens metadata URLs to all three credentials and says what a self-served key cannot prove

    Draft · · IETF Datatracker · Arndt Schwenkschuster et al. · Jun 15, 2026

  20. Raut's revised agent draft keeps token exchange out of the model's reach, and concedes the token carries no evidence of it

    Draft · · IETF Datatracker · Ashay Raut · May 22, 2026

  21. An unadopted AGTP draft would let gateways refuse an agent's out-of-scope requests from its TLS certificate

    Draft · · IETF Datatracker · Chris Hood, Nomotic, Inc. · Oct 30, 2026

  22. Give an agent access for the task and take it back afterwards, four voices say, and four say people approving each request cannot keep up

    Field signals · · 8 voices from the field · Feb 2026 to Oct 2026

  23. IETF workload practices draft describes single-audience JWTs in place of stored secrets and discourages token exchange

    Draft · · IETF Datatracker · Arndt Schwenkschuster et al. · Sep 22, 2026

  24. WIMSE credentials -03: a valid signature proves trust-domain membership, not which workload is calling

    Draft · · IETF Datatracker · Brian Campbell et al. · Jul 2, 2026

  25. AGTP checks that an agent ID is signed; binding it to the connection is an add-on in the author's patent notice

    Draft · · IETF Datatracker · Chris Hood, Nomotic, Inc. · Jun 28, 2026

  26. Agent delegation records not anchored in trusted time become deniable once timestamps can be forged

    Draft · · IETF Datatracker · V.K. Uppalapati · Sep 29, 2026

  27. 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

  28. 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

  29. 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

  30. OAuth scopes grant agents far more than the task needs, four voices say, and authorization handed from agent to agent has no settled pattern yet

    Field signals · · 7 voices from the field · Oct 2025 to Oct 2026

  31. LinkedIn, Uber and State Farm already run SPIRE for workload identity, and each plans to extend it

    Field signals · · 12 voices from the field · Nov 2024 to Oct 2026

  32. A cross-domain transaction-token grant now names AI agents as a use and keeps the call chain at home

    Draft · · IETF Datatracker · George Fletcher et al. · Sep 10, 2026

  33. IBM Research puts agent identity in a SPIFFE and OAuth sidecar; Maia Iyer calls authorization policy still open

    Podcast · · Kubernetes Bytes · Your AI Agent Has Too Much Access | Fixing Agent Identity · Sep 15, 2026

  34. The IETF names workloads with a URI on the owner's domain; revision -04 pins down how two names are compared

    Draft · · IETF Datatracker · Yaroslav Rosomakho et al. · Oct 23, 2026

  35. At OWASP, one poisoned memory entry made a demo agent send a Q4 report to an attacker server

    Video · · OWASP Foundation · Recorded Jun 2026 · Uploaded October 21, 2026

  36. 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

  37. 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

  38. An individual IETF draft takes a workload's role and group out of its name and puts them in separate metadata

    Draft · · IETF Datatracker · Shlomo Zalman Heigh et al. · Oct 16, 2026

  39. Six companies' agent-auth draft, since adopted by the WIMSE group, keeps credentials out of the model's reach

    Draft · · IETF Internet-Draft (individual submission, since adopted by the WIMSE working group) · Pieter Kasselman et al. · Jul 6, 2026

  40. An OAuth profile for AI agents makes every hand-off cut privileges, but a shorter expiry alone counts as a cut

    Draft · · IETF Datatracker · Angel Cruz · Feb 7, 2026

  41. Anthropic asks for per-agent identity; its guide leaves Claude Code sub-agent credentials unspecified

    Article · · Anthropic (Claude) · Anthropic · May 27, 2026

  42. 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

  43. An end-to-end encrypted draft would hide an AI agent's task from the broker, but not who sent it or to whom

    Draft · · IETF Datatracker · Hamdy Atakora, Imara Labs · Sep 21, 2026

  44. An attested OAuth client is one somebody vouched for; revision 12 says to trust its claims only as far as the voucher

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

  45. 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

  46. Munoz's agent Permit is signed before the action runs, but the profile does not make the action wait for it

    Draft · · IETF Datatracker · Christian Munoz, Keel API, Inc. · Jul 19, 2026

  47. The PEDIGREE draft has verifiers re-check every ancestor of an agent's token, without saying how to fetch them

    Draft · · IETF Datatracker · Karthik Rampalli, Glyphzero Labs · Apr 25, 2026

  48. 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

  49. Sharif's profile writes an AI agent's rights to pay and delete into an identity certificate valid up to a year

    Draft · · IETF Datatracker · Raza Sharif, CyberSecAI · Oct 2, 2026

  50. OWASP's agent Top 10 blames identity systems built for people, and does not say how it chose or ranked its risks

    Article · · OWASP GenAI Security Project · John Sotiropoulos et al. · Dec 9, 2025

  51. NIST's agent identity work starts from OAuth, SPIFFE and SCIM and defers agents from outside the enterprise

    Article · · NIST National Cybersecurity Center of Excellence · Harold Booth et al. · Feb 5, 2026

  52. NIST and CISA advise one-hour tokens, 90-day keys and hardware key storage; AI agents get one paragraph

    Article · · NIST (CSRC) · Ryan Galluzzo et al. · Sep 15, 2026

  53. Saviynt's strategy head calls agent discovery a commodity and pitches Zuma for runtime control

    Podcast · · The Non-Human & AI Identity Podcast · Ep # 14 - Turning an AI Breach into a Case Study · Sep 15, 2026

  54. AAuth's bootstrap guide makes device attestation optional and lets a new key mean a new agent

    Draft · · IETF Datatracker · Dick Hardt, Hellō · Sep 28, 2026

  55. A draft ties what each AI agent may read to what it may do, and still can't take a grant back

    Draft · · IETF Datatracker · Hamdy Atakora · Sep 21, 2026

  56. 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

  57. PayPal moved MCP tool permission checks into the server, and prompt-injected agents stopped getting through

    Paper · · arXiv · Huan Li et al. · Sep 18, 2026

  58. 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

  59. Coetzee's agent token is good for one declared call, and the hard part is hashing that call the same way twice

    Draft · · IETF Datatracker · Rudolf Jacobus Coetzee · Jul 19, 2026

  60. A SANS instructor's broker vaults agents' real keys, but its third delivery model breaks the abstract's promise

    Draft · · IETF Datatracker · Kenneth G. Hartman · Mar 29, 2026

  61. Eight panels agree an agent acts for a person, never as that person. When agents hand work to other agents, nobody had an answer.

    Conference · · Las Vegas · June 2026, a one-day summit track inside Identiverse 2026; the eight panels were published as podcast episodes between 18 August and 7 September 2026

  62. 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

  63. Four vendors now say the access review is finished for agents. This one adds that agents will approve agents.

    Podcast · · The Non-Human & AI Identity Podcast · Ep # 15 - Identity is the 3rd Generation of Security · Sep 22, 2026

  64. The IETF's workload-identity architecture gives AI agents one subsection, calling an agent a delegated workload

    Draft · · IETF Datatracker · Joseph Salowey et al. · Jul 6, 2026

  65. 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

  66. Two ways to give an AI agent an OAuth identity clashed in April; a draft now lets them work together

    Podcast · · Hybrid Identity Protection Podcast · Securing Non-Human Identities in the Age of Agentic AI with Sarah Cecchetti, Director of Product Management at Semperis · Apr 28, 2026

  67. Okta's answer to agent sprawl: put the identity provider between every agent and every app

    Podcast · · Cloud Security Podcast · Who Governs Your AI Agents? Identity, Offboarding & Open Standards · Jul 2, 2026

  68. Keeping an agent's delegation chain is optional, and room in the HTTP header decides whether it survives

    Podcast · · The Non-Human & AI Identity Podcast · NHI & AI Identity Summit : The New NHI Risk Reality · Aug 26, 2026

  69. Highflame's open-source ZeroID narrows an agent's token at every hop, a step we did not find in Uber's write-up

    Article · · Highflame (GitHub) · Sep 25, 2026

  70. The agent risk two CISOs fear is manipulation, not a stolen credential

    Podcast · · Defense in Depth · Identity and Access Management (IAM) in an Agentic AI World · Jul 23, 2026

  71. Agent identifiers are split across JWT-based, SPIFFE, WIMSE and proprietary schemes, an identity lead says

    Podcast · · Identity Decoded · Winning in securing agentic identity: Start with one team, one use case · Aug 12, 2026

  72. Canva sandboxes bounded conversion jobs; broad workforce agents pose a different access problem

    Podcast · · AI Security Podcast · Why I Dont Trust Your AI Agent | Kane Narraway, Canva · Aug 20, 2026

  73. IPC passes a delegation chain between services; its signature proves only the last signer

    Draft · · IETF Datatracker · Ravi Kant Sharma, Ericsson · Jul 23, 2026

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

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

  75. A June OAuth draft logs each hand-off between AI agents, but the per-hop limits the act claim lacks are optional

    Draft · · IETF Datatracker · Dapeng Liu et al. · Jun 8, 2026

  76. Google's agent registry went GA in June while Okta, SailPoint and Ping pulled agents into their own lists

    Release tracker · · 11 vendors' release notes · Jun 2026

  77. 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

  78. 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

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

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

  80. Three vendors put MCP controls in three different places in May, and only Auth0's was GA

    Release tracker · · 10 vendors' release notes · May 2026

  81. 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

  82. An agent certificate draft ties access to a trust score set by the issuing CA, usually the agent operator's own

    Draft · · IETF Datatracker · Raza Sharif, CyberSecAI · Apr 10, 2026

  83. AITH claims 4.7 million agent checks a second, but the test simulated its post-quantum signatures

    Paper · · arXiv · Zhaoliang Chen · Apr 9, 2026

  84. 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

  85. Outside Azure, Argo CD uses stored secrets for repositories; a March plan for per-project identity is still open

    Video · · CNCF [Cloud Native Computing Foundation] · Uploaded March 24, 2026

  86. A KubeCon demo moves MCP servers' OAuth checks into a gateway, safe only if nothing can reach a server directly

    Video · · CNCF [Cloud Native Computing Foundation] · Uploaded March 24, 2026

  87. An agent identity draft signs every tool call, but never finished saying whom the agent answers to

    Draft · · IETF Datatracker · James Cao et al. · Mar 16, 2026

  88. 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

  89. A draft binds workload credentials to their host via a local agent, and leaves that agent's compromise undecided

    Draft · · IETF Datatracker · Ram Krishnan et al. · Feb 26, 2026

  90. MACAW Security's trust layer signs every agent call, but what stopped its prompt-injection test was a policy

    Paper · · arXiv · Mohan Rajagopalan et al. · Feb 11, 2026

  91. ACME's device-attestation draft now advises leaving device IDs out of certificates and drops a key-export rule

    Draft · · IETF Datatracker · Brandon Weeks et al. · Jan 29, 2026

  92. Uber reports authorizing over 2 million service-to-service connections a second after staged rollout

    Video · · CNCF [Cloud Native Computing Foundation] · Uploaded November 24, 2025

  93. State Farm moved SPIRE from many token issuers to one, then hit an AWS key quota and a 15-character name limit

    Video · · CNCF [Cloud Native Computing Foundation] · Nov 24, 2025

  94. A genuine TPM does not prove the server is LinkedIn's, so its SPIRE first checks the switch port and cabinet

    Video · · CNCF [Cloud Native Computing Foundation] · Uploaded November 24, 2025

  95. 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

  96. WIMSE leaves its application proof choice open, with different integrity coverage

    Draft · · IETF Datatracker · Brian Campbell et al. · Oct 16, 2025

  97. SPIFFE can authenticate an OAuth client; trust and registration still come first

    Draft · · IETF Datatracker · Arndt Schwenkschuster et al. · Oct 3, 2025

  98. 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

  99. Akeyless's CEO called just-in-time credentials the current focus and secretless an ideal that will take time

    Podcast · · The Non-Human & AI Identity Podcast · The NHI Global Summit - Fireside Chat:Managing the full-lifecycle of Secrets & Non-Human Identities · First aired February 27, 2025 · Sep 4, 2025

  100. Uber had workload identity on a quarter-million nodes two years before agents; certificates were the easy part

    Article · · Uber Engineering · Andrew Moore et al. · Nov 9, 2023

  101. 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

In our columns

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

    Column · Oct 15, 2026

  2. An 'attested' agent is one whose key somebody vouched for. Ask what they checked

    Column · Oct 6, 2026

  3. Give new agents platform identity now; stored secrets take years to clean up

    Column · Oct 4, 2026

  4. The one agent delegation chain we've seen running was built in-house; plan to build yours

    Column · Sep 22, 2026

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

    Column · Sep 17, 2026

Related terms

What appears alongside SPIFFE in the same briefings.

OAuth SPIFFE Client AuthenticationSPIREWIMSELinkedInZeroIDCross App AccessSecretlessWIT

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.