A map of the standards and regulations that govern AI-agent identity, authorisation and accountability.

Standards & regulation tracker

Know Your Agent (KYA) is being built in public, by many hands, in several places at once. This page tracks the standards bodies, regulators and payment schemes that matter for knowing who an agent is, who is accountable for it, and whether it may act — and records where each effort stands, dated.

Read this as an evergreen map, not a news feed. Statuses and dates change; the questions do not. For the underlying concepts, start with What is KYA? and How to verify an agent.

State of play, as of 8 October 2026

There is still no adopted standard for agent identity. No IETF RFC, no ISO/IEC standard and no regulation yet defines a complete, certifiable answer to “who is this agent, on whose authority does it act, and how is it revoked”. What exists is a set of overlapping workstreams converging on the same primitives — cryptographic identity, scoped delegation, signed proof of intent, and a revocable trust anchor — plus payment schemes moving faster than the standards bodies because money is where the risk concentrates.

The active venues are:

  • NIST’s National Cybersecurity Center of Excellence (NCCoE) is turning a Feb 2026 concept paper on Software and AI Agent Identity and Authorization into a practice guide (an SP 1800-series document), using OAuth 2.0, SPIFFE/SPIRE and MCP.
  • The IETF is the main technical venue. Its WIMSE working group has adopted AI Identity Management System (`draft-ietf-wimse-aims`) as a working-group item and is standardising workload-identity practice; the Web Bot Auth effort would let automated clients sign HTTP requests so servers can tell legitimate agents from hostile bots.
  • The FIDO Alliance now houses the payments-side standardisation: it took in Google’s Agent Payments Protocol (AP2) and Mastercard’s Verifiable Intent in April 2026 and runs Agentic Authentication and Payments technical working groups.
  • The W3C supplies the credential data model (Verifiable Credentials 2.0) that agent credentials reuse; the OpenID Foundation has published a strategic agenda for agentic AI identity.
  • Payment schemes and platforms (Visa, Mastercard, OpenAI/Stripe, Google, Coinbase) have shipped agent-payment protocols ahead of formal standards.
  • Regulators have not written agent-identity rules as such. The EU AI Act imposes transparency and record-keeping duties on AI systems; MAS in Singapore has published a runtime-governance framework for AI agents in finance; AML rules still apply to the principal, not the agent.

The practical takeaway: the building blocks are mature (OAuth 2.1, mTLS, X.509, SPIFFE, verifiable credentials); the assembled practice is not. Any vendor claiming a finished “KYA standard” should be asked which of those blocks it uses and how it handles revocation and delegation.

The tracker

Body / publisherItemStatusKey datesWhat it means for KYAPrimary link
NIST NCCoESoftware and AI Agent Identity and Authorization (project)Concept paper published; comment period closed; practice guide in progressConcept paper Feb 2026; comments closed 2 Apr 2026; 600+ comments; outcome = SP 1800-series practice guideThe most authoritative US government effort to define how identification, authentication and authorisation apply to agents. Uses OAuth 2.0, SPIFFE/SPIRE and MCP as reference technologies.https://pages.nist.gov/nccoe-ai-identity/ ; https://www.nccoe.nist.gov/news-insights/new-concept-paper-identity-and-authority-software-agents
IETF WIMSE (Workload Identity in Multi System Environments)AI Identity Management System (`draft-ietf-wimse-aims-00`)Working-group Internet-Draft (Informational), adopted 15 Sep 2026Adopted 2026-09-15; drafts expire on their own scheduleThe IETF’s first working-group document on agent authentication/authorisation. Frames agents as delegated workloads and proposes best practices over OAuth, SPIFFE and WIMSE. Not an RFC.https://datatracker.ietf.org/wg/wimse/documents/
IETF WIMSEArchitecture and workload-credential drafts: `draft-ietf-wimse-arch`, `-workload-creds`, `-http-signature`, `-wpt`, `-identifier`, `-mutual-tls`; Workload Identity Practices (`-workload-identity-practices`)Active working-group drafts; Practices submitted to the IESG (Informational)Ongoing; Practices in IESG reviewThe substrate agent identity sits on: how workloads prove identity across systems. Directly reusable for agents.https://datatracker.ietf.org/wg/wimse/documents/
IETF WIMSE (related individual drafts)`draft-ni-wimse-ai-agent-identity`, `draft-gilda-wimse-agent-audit-record`, `draft-asor-wimse-agent-delegation-chain`Individual submissions (not adopted)2026Early proposals for agent identity, agent audit records, and delegation chains. Indicative of direction; not standards.https://datatracker.ietf.org/wg/wimse/documents/
IETF Web Bot AuthHTTP Message Signatures for automated traffic (`draft-ietf-webbotauth-httpsig-protocol`), built on HTTP Message Signatures (RFC 9421)Working-group Internet-DraftOngoing through 2026Lets an automated client cryptographically sign outbound HTTP requests so a server can verify its identity — a lightweight agent/bot identity for the open web. Deployed early by Cloudflare and Akamai.https://datatracker.ietf.org/doc/draft-ietf-webbotauth-httpsig-protocol/
W3CVerifiable Credentials Data Model 2.0W3C Recommendation15 May 2025The issuer/holder/verifier data model that agent credentials and payment mandates reuse. Underpins “verifiable” agent claims.https://www.w3.org/TR/vc-data-model-2.0/
IETF OAuth WGSD-JWT-based Verifiable Credentials (SD-JWT VC)Active IETF workOngoingSelective-disclosure credential format increasingly used for agent and payment credentials.https://datatracker.ietf.org/doc/draft-ietf-oauth-sd-jwt-vc/
OpenID FoundationIdentity Management for Agentic AI (whitepaper, Oct 2025)Published strategy/whitepaperOct 2025; workstream ongoingArgues today’s standards secure simple agents but gap on recursive delegation and cross-domain trust; recommends interoperable profiles of existing standards.https://openid.net/new-whitepaper-tackles-ai-agent-identity-challenges/
FIDO AllianceAgentic Authentication Technical Working Group & Payments Technical Working GroupWorking groups formed and runningFormed 28 Apr 2026The venue where agent-payment credentials are being standardised; received Google’s AP2 and Mastercard’s Verifiable Intent as initial contributions.https://fidoalliance.org/fido-alliance-to-develop-standards-for-trusted-ai-agent-interactions/
Google / FIDOAgent Payments Protocol (AP2)Open protocol; v0.2 released and donated to FIDO; standardisation continues in FIDO TWGsAnnounced 16 Sep 2025; v0.2 + FIDO donation 28 Apr 2026Defines Verifiable Digital Credentials (Checkout and Payment Mandates, Open/Closed) so a merchant can verify a user’s intent, not just a credential. The clearest worked model of agent identity inside a transaction.https://ap2-protocol.org/
VisaTrusted Agent Protocol (TAP) / Visa Intelligent CommerceOpen framework, documented on Visa Developer CenterAnnounced Dec 2025; docs updated 2026Lets merchants distinguish legitimate AI agents from malicious bots, and validate an agent-entered payment credential, using existing web infrastructure.https://developer.visa.com/capabilities/trusted-agent-protocol/overview
MastercardAgent Pay / Agentic Tokens; Verifiable Intent frameworkLive programme built on network tokenisation (MDES)Announced 29 Apr 2025; contributed to FIDO 2026Registers, verifies and recognises agentic transactions; Agentic Tokens carry the user’s permissions and limits. Co-developed Verifiable Intent with Google.https://www.mastercard.com/us/en/business/artificial-intelligence/mastercard-agent-pay.html
OpenAI / StripeAgentic Commerce Protocol (ACP) + Delegated Payment Spec / Shared Payment TokenOpen protocol; Stripe is the first compatible PSPSpec from Sep 2025; evolving 2026Lets an agent complete merchant checkout with a scoped, single-use payment credential. OpenAI is not the merchant of record.https://www.agenticcommerce.dev/
GoogleUniversal Commerce Protocol (UCP)Open-source standard; published Jan 2026Jan 2026; updated Mar 2026Orchestrates discovery, checkout, payment and fulfilment for agentic commerce; AP2 is its payments extension.https://developers.google.com/merchant/ucp
Coinbase / x402 Foundationx402Open payment standard over HTTP 402Launched 2025; foundation announced 2026Instant stablecoin payments over HTTP so an agent can pay per request; an alternative rail to card-based mandates.https://x402.org/
Monetary Authority of Singapore (MAS)Safeguards for Agentic Finance at Runtime (SAFR)Industry white paper v1.03 Jul 2026Inserts a governance checkpoint between an agent’s proposed action and its execution in financial services; defines which agents may act for which principals and to what delegation depth.https://www.mas.gov.sg/publications/monographs-or-information-paper/2026/safeguards-for-agentic-finance-at-runtime
EUAI Act (Regulation (EU) 2024/1689) and the “AI Omnibus” (Regulation (EU) 2026/1744)In force; staged applicationAI Act in force 1 Aug 2024; applicable 2 Aug 2026; Omnibus in force 27 Jul 2026; Annex III high-risk from 2 Dec 2027; Annex I from 2 Aug 2028Art. 50 transparency (tell people they are interacting with an AI, incl. agents) and Art. 12 record-keeping give a transparency and logging duty, not identity verification.https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
EUeIDAS 2.0 — Regulation (EU) 2024/1183 (European Digital Identity Framework)In force; wallets and implementing acts rolling outIn force 2024; wallet rollout tied to implementing actsCreates a state-backed verifiable-credential ecosystem (EU Digital Identity Wallets) that agent credentials and delegation can build on.https://digital-strategy.ec.europa.eu/en/policies/eudi-regulation
ISO / IECISO/IEC 42001:2023 — AI management systems (AIMS), part of the 42 000 familyPublished international standard; certifiablePublished Dec 2023Gives organisations a certifiable governance system for AI, within which agent identity and access controls are a control to be managed. Does not define agent identity itself.https://www.iso.org/standard/42001
FATFThe FATF Recommendations (Recommendation 10, customer due diligence)In force; no agent-specific rule yetOngoingAML/KYC duties attach to the principal (the human or legal person). An agent does not remove the principal’s due-diligence obligation; agent-specific AML guidance has not been issued.https://www.fatf-gafi.org/en/publications/Fatfrecommendations/Fatf-recommendations.html

What to watch next

  1. The NCCoE practice guide. When NIST’s SP 1800-series guide lands, it becomes the reference implementation most enterprises will benchmark against. Watch which of its focus areas (identity, authorisation, logging, lifecycle) it treats as normative.
  2. Whether `draft-ietf-wimse-aims` progresses. Adoption is not standardisation. The signal to watch is whether WIMSE milestones it and whether related individual drafts (agent identity, audit record, delegation chain) are adopted into the group.
  3. FIDO’s first agentic specifications. FIDO has the members and the process; the test is whether it can converge AP2, Verifiable Intent and rival schemes into interoperable profiles rather than parallel ones.
  4. The Web Bot Auth deployment curve. Signature-based bot/agent authentication becoming default at internet scale would give the open web a working, if narrow, agent-identity primitive.
  5. AI Act enforcement. With Annex III high-risk duties deferred to Dec 2027, near-term attention is on Art. 50 transparency from Aug 2026 and the guidance around it — a duty about disclosure, not identity.
  6. Payment-scheme consolidation. Multiple card, wallet and crypto rails are competing (AP2, TAP, Agent Pay, ACP, UCP, x402). Convergence — or a durable split — will shape what “verifiable intent” means commercially.
  7. Revocation and delegation at scale. The unsolved parts are the same everywhere: proving a sub-agent’s authority can only narrow, and stopping a compromised agent in seconds across many relying parties.

A note on method

Entries above are drawn from the issuing body’s own pages wherever possible, with the access date below. Where a fact is a status rather than a specification (for example, that a comment period closed, or that a working group was formed), it is attributed to the body’s own announcement or page. Anything we could not verify against a primary source has been left out of this tracker rather than softened into a vague claim.

Primary sources

Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.