Talk to usPricingPartnersEnglish · UK · GBP

Trust

Neutral by knowing less.

WorldAuth can only be the control point between parties if it is not quietly becoming one of them. So it holds operational trust metadata — not the world’s identities or activity.

01 Three principles

What WorldAuth refuses to hold.

These are architectural exclusions, not configuration options.

No global ledger

Decisions are receipted where they belong

There is no central record of who did what. Full Receipts, sensitive evidence and raw transaction content remain customer- and relying-party-controlled by default.

No reputation economy

Authority comes from a Mandate, not a score

WorldAuth calculates no universal reputation or behaviour score for any person, organisation or agent. We do not rank actors.

Assurance stays behind the firewall

Evidence informs; it never travels as authority

Identity and evidence can inform whether a Mandate is granted. They never pool into a profile and never travel with the Key.

02 Security

Failure must never create authority.

WorldAuth’s security claims come from architecture, not adjectives. This section states what the design guarantees, what depends on a runtime that is still being built, and how to report a vulnerability.

Fail closed

Uncertainty is not permission

Malformed, expired, replayed, unbound or unverifiable authority is non-authorising. Network, cache, parser, status, clock and policy failure map to a non-authorising outcome — never to ALLOW.

Binding

Possession alone grants nothing

A presentation is bound to an audience, a resource, an action and a server-issued challenge, and carries proof of possession over exactly those bytes. A captured presentation does not transfer.

Attenuation

Delegation only narrows

Every hop is checked in every dimension, and the chain is rebuilt from its root. Widening, chain splicing and unrelated-parent substitution are rejected.

Separation

A decision is not an outcome

Deciding and doing are separate records. Decision evidence cannot be presented as proof that something was executed.

Firewall

Assurance is never authority

Identity, KYA, controls, insurance and provenance evidence inform a grant. They cannot widen a Mandate.

Minimalism

As little data as possible

Sensitive evidence and Receipts stay customer-controlled by default. There is no universal central activity ledger.

03 Honest limits

What is proven, and what is not.

The published contract and its conformance suite prove properties of the artefacts. They do not prove that a Gate enforces them at request time, because Gate Core does not exist yet.

PropertyEvidence todayStatus
Object integrityReal signature verification over canonical bytes, with tamper and forgery vectors.Demonstrated
Proof of possessionOne normative signing input, with wrong-key, challenge-substitution and absent-proof vectors.Demonstrated
Delegation chainA two-hop chain plus splice, substitution and hop-count vectors.Demonstrated
Replay defenceRequires a Gate to enforce single-use challenge nonces.Runtime — not yet
Status freshness & revocationRequires a Gate to fail closed without fresh status.Runtime — not yet
Depletion safetyRequires a coordinator to serialise reservations.Runtime — not yet
Tenant isolation, DoS resistanceControl-plane and operational properties outside the object model.Runtime — not yet
Signing-key custodySoftware custody today; HSM/KMS custody is on the roadmap and named as such.Roadmap
Post-quantum sealingEd25519 today; hybrid classical and post-quantum signature suites on the roadmap ahead of the 2027 compliance gates.Roadmap

WorldAuth makes no claim of being unhackable, “military-grade” or absolutely secure.

04 Coordinated disclosure

If you find something, tell us first.

If you believe you have found a security issue in WorldAuth — the protocol contract, this site, or anything we operate — please report it privately before publishing.

01Wheresecurity@worldauth.com — a monitored private channel, also published in /.well-known/security.txt. Include enough detail to reproduce, and tell us if you intend to publish.Write to us
02WhenWe aim to acknowledge within three working days and to give an assessment within ten. Serious issues are fixed or mitigated as a priority and the outcome is published once closed.3 · 10 working days
03What we askGive us a reasonable opportunity to remediate before publishing, and do not access, alter or retain data that is not yours. Good-faith research reported through this channel is welcome.Good faith

We do not currently operate a paid bug-bounty programme, and we will not claim one until it exists.

05 Privacy posture

Designed to know as little as possible.

WorldAuth is authority infrastructure, not a data business.

Keys

Authority, not identity

A Key reveals only what the Gate needs to decide. No identity documents, KYC, biometrics or reputation travel with it.

Receipts

What was authorised

A Receipt records what was authorised — never a person’s documents, face or history. Verification does not phone home.

Residency

By deployment

Where a Gate runs is where its decisions and evidence live. The Sovereign profile keeps trust material, policy, coordination and evidence inside your boundary.

What this website itself collects is set out in the privacy notice. Data questions and rights requests: privacy@worldauth.com.

06 Responsible deployment

A Gate must never become a means of exclusion.

Principles that bind every deployment, including our own.

  • Step-up and human fallback are mandatory wherever a decision affects a person’s access to essential services.Human in the loop
  • Availability, language and display currency are presentation only and never change authority or entitlement.Presentation ≠ authority
  • Public decisions are exactly ALLOW, STEP-UP and DENY; a Gate never invents a fourth outcome to route around a person.Three decisions
  • Local deployment can support a regulatory position; it does not by itself make an operator compliant with any regime, and we do not say otherwise.No compliance by proxy

Status Protocol v0.1 in draft · reference implementation in development · not yet in production use

Questions a security review would ask?
Ask them of us directly.