ĀML™ — ĀRU Meaning Language™

The accountability layer between AI and the human interface.

AI can propose an interface. ĀML can make its meaning, policy, authority, and evidence inspectable before and after rendering.

HTML tells the browser what to display. ĀML tells the system why it deserves to be displayed.

ĀML is a working research prototype for meaning-native, policy-aware, accountable AI interfaces. It does not require replacing HTML, React, or existing frontend stacks. It can sit between machine intent and human-facing output as an AI Interface Firewall™.

Release status: v1.3.0 remains the stable package/CLI/capability contract. v1.4.0-rc.2 is the current GitHub prerelease snapshot of the broader architecture on main. It is not a claim that a stable v1.4 registry package has already shipped.

Three ways in

TRY IT

Open the proof and change one number:

https://aruintelligence.github.io/aml-core/proof.html?attention=5&restoration=1&lang=en

READ IT

Start with the public reading room:

ĀML Publications →

Recommended first reads:

• Start Here — ĀML in 5 minutes
• ĀML in one page
• Why AI-generated UI needs a firewall
• A critic's guide to ĀML
• Enterprise buyer brief

INTEGRATE IT

• Developer integration brief
• 10-minute reproduction
• 30-minute enterprise pilot
• Verify AML without trusting AML

Start with proof

1. Open the live proof: https://aruintelligence.github.io/aml-core/proof.html?attention=5&restoration=1&lang=en
2. Watch the same interface decision show SUPPRESS.
3. Change restoration_value from 1 to 5.
4. Watch it become ALLOW.
5. Copy the exact proof URL and send it to someone else.
6. Reproduce the deterministic receipt locally with demos/undeniable-proof/.

Prototype rule:

render_allowed = restoration_value >= attention_cost

The scores are declared/model inputs in this prototype, not claimed objective measurements of human cognition or wellbeing.

Start in 60 seconds

• Publications: PUBLICATIONS.md
• Proof: https://aruintelligence.github.io/aml-core/proof.html
• Decision gallery: https://aruintelligence.github.io/aml-core/gallery.html
• Zero-install <aml-gate> demo: https://aruintelligence.github.io/aml-core/aml-gate-demo.html
• Existing-HTML bridge demo: https://aruintelligence.github.io/aml-core/dom-gate-demo.html
• Offline single-file proof: https://aruintelligence.github.io/aml-core/offline-proof.html
• Try AML: https://aruintelligence.github.io/aml-core/playground.html
• View Meaning™: https://aruintelligence.github.io/aml-core/view-meaning.html
• Verify Official AML: https://aruintelligence.github.io/aml-core/official-verify.html
• Why AML Now: docs/WHY_AML_NOW.md
• 10-minute reproduction: docs/TRY_AML_10_MINUTES.md
• Out-of-the-box adoption: docs/OUT_OF_THE_BOX.md
• 30-minute enterprise pilot: pilots/enterprise-30min/
• Full JavaScript API: API.md
• v1.4 RC2 notes: RELEASE_NOTES_1.4.0_RC2.md

Zero-install browser entry

<script type="module">
  import { compileSourceBrowser } from "https://aruintelligence.github.io/aml-core/aml-browser.js";

  const result = compileSourceBrowser(`transmission "demo" {
    message "welcome" {
      purpose: "Explain the interface"
      attention_cost: 1
      restoration_value: 3
    }
  }`);

  console.log(result.amt);
  console.log(result.renderDecisions);
</script>

One wrapper around existing HTML

<script type="module" src="https://aruintelligence.github.io/aml-core/aml-gate.js"></script>

<aml-gate
  purpose="Create urgency"
  attention-cost="5"
  restoration-value="1">
  <button>Act now</button>
</aml-gate>

See docs/AML_GATE_ELEMENT.md.

Or keep the DOM and add three attributes

<script type="module" src="https://aruintelligence.github.io/aml-core/aml-dom-gate.js"></script>

<div
  data-aml-purpose="Create urgency"
  data-aml-attention-cost="5"
  data-aml-restoration-value="1">
  Offer expires soon.
</div>

See docs/HTML_BRIDGE.md.

The two browser bridges are deliberately narrow adoption surfaces. They do not claim to contain the full runtime policy, consent, privacy, accessibility, receipt, or trust stack.

Proof can travel

• Exact score state is encoded in the proof URL.
• The proof URL can also carry language and declared purpose.
• Live proof UI currently supports English, Spanish, Portuguese (Brazil), Hindi, Arabic, Chinese, French, and German.
• Arabic switches to RTL.
• proof-card.html provides an embeddable proof surface.
• proof-badge.svg means only proof available; it is not a certification badge.
• proof-manifest.json and proof-links.json make proof machine-readable.
• .well-known/aml.json is an experimental reference-project discovery convention, not a ratified Internet standard.

The core execution model

flowchart LR
    A[AI / app intent] --> B[ĀML Interface Firewall]
    B --> C[Intent → AML]
    C --> D[AST + Abstract Meaning Tree]
    D --> E[Semantic + accessibility analysis]
    E --> F[Policy matrix / consensus]
    F --> G[Consent + privacy + attention + authority]
    G --> H[Render decision]
    H --> I[HTML / existing UI]
    H --> J[Audit stream + receipt]
    J --> K[Proof / provenance / federation]
    K --> L[Independent verification / View Meaning]

Loading

What AML can do today

Meaning and policy

• lexer/parser + AST + Abstract Meaning Tree
• semantic diffs and semantic risk scoring
• policy diffs and multi-policy matrices
• policy consensus with explicit dissent
• pluggable policies and user/organization profiles
• deterministic machine-intent → AML generation

Human-context controls

• consent-aware and privacy-aware policies
• expiring/revocable consent ledger
• reduced-motion, contrast, keyboard, text-alternative, and cognitive-load audits
• cumulative session attention accounting

Accountability evidence

• execution receipts
• Ed25519-signed receipts
• signed data-only policy packs
• SHA-256 runtime audit streams
• signed audit checkpoints
• execution provenance graphs
• Merkle-batched receipt inclusion proofs
• Proof-Carrying Interface™ manifests

Federated trust

• capability negotiation
• portable policy passports
• content-addressed artifact bundles
• selective-disclosure commitments
• versioned aml-wire/1 envelopes
• replay protection
• causal execution graphs
• canonical JSON + golden protocol vectors
• trust delegation
• M-of-N threshold authorization
• bounded signed capability tokens
• transparency logs
• revocation registries

Mainstream adoption surfaces

• public publication library for developers, designers, AI teams, security/privacy, research, enterprise, and media
• shareable exact-state proof URLs
• eight-language live proof with Arabic RTL
• embeddable proof card + proof-available badge
• decision gallery with five ALLOW / five SUPPRESS examples
• zero-install <aml-gate> custom element
• existing-HTML data-aml-* bridge
• single-file offline proof demonstrator
• machine-readable proof discovery
• AI Interface Firewall™
• React-compatible accountable UI adapter
• plain JavaScript / React / Next.js starters
• dependency-free HTTP evaluation service
• Meaning Gate™ GitHub Action
• View Meaning™ browser inspector
• privacy-minimal Manifest V3 View Meaning extension prototype
• 30-minute enterprise pilot kit

AI Interface Firewall™

import { createInterfaceFirewall } from "./index.js";

const firewall = createInterfaceFirewall({ profile: "human_first" });
const result = firewall.enforce({
  transmission: "pricing_assistant",
  nodes: [{
    type: "message",
    identifier: "pricing",
    properties: {
      purpose: "Explain pricing clearly",
      content: "Simple pricing",
      attention_cost: 1,
      restoration_value: 2
    }
  }]
});

if (result.allowed) render(result.html);

The result can carry policy decisions, accessibility analysis, provenance, attention accounting, audit state, and a verifiable execution receipt.

View Meaning™

The web gave developers View Source. ĀML explores View Meaning™.

import { viewMeaning } from "./index.js";
const report = viewMeaning(receipt);

A View Meaning report can expose declared purpose, policy/profile, consent/privacy/accessibility context, attention/restoration inputs, allow/suppress outcome, rationale, and receipt integrity.

Browser inspector: https://aruintelligence.github.io/aml-core/view-meaning.html

The extension prototype under extensions/view-meaning/ performs user-invoked active-tab inspection, recomputes receipt integrity locally, and can verify official-brand credentials against the canonical public ĀRU trust-root registry.

Meaning Gate™

- uses: aruintelligence/aml-core/actions/meaning-gate@main
  with:
    before-file: before.aml
    after-file: after.aml
    before-policy: calm_default
    after-policy: human_first

The gate can surface or block high-risk semantic changes, new personal-data collection, consent changes, attention/restoration changes, and policy regressions.

Run AML as a service

npm run serve
# http://127.0.0.1:8787

Reference HTTP endpoints and the official-verification contract are documented in protocol/aml-http.openapi.yaml.

The reference service does not replace production authentication, authorization, TLS, rate limiting, logging, network controls, or secure secret management.

Open technology. Controlled official identity.

The covered software is available under the MIT License. Official ĀML™ / ĀRU™ branding, logos, compatibility branding, certification-style identity, endorsement, OEM/co-branding, and related reserved brand rights are separate from the software license.

Technical conformance is independently testable. It does not automatically grant official authorization, endorsement, partnership, certification, or trademark rights.

Official resources:

• TRADEMARKS.md
• OFFICIAL_MARKS.json
• OFFICIAL_AUTHORIZATIONS.json
• BRAND_TRUST_ROOTS.json
• COMMERCIAL.md
• RFC 0011 — Official Brand Authorization
• Trademark registration plan

Commercial, OEM, enterprise, and strategic inquiries: Office@aruintelligence.com

Cryptographically verifiable official AML authorization

After an appropriate written agreement, ĀRU can issue a scoped aml-brand-authorization/1 credential.

aml sign-brand-authorization authorization.json aru-private-key.pem credential.json
aml verify-brand-authorization credential.json revocation-registry.json
aml-brand-verify credential.json BRAND_TRUST_ROOTS.json

A credential can be cryptographically valid while still not being an official ĀRU authorization. Official verification additionally requires its signer fingerprint to be active and non-revoked in the canonical public trust-root registry.

Current production public trust root:

• key id: aru-aml-brand-prod-2026-09-08-01
• public SHA-256 fingerprint: eda0184568cb2110add5130d2a9fffaf53a77e0f2be311be414e7912ed69997c
• public key: keys/aru-aml-brand-prod-2026-09-08-01-public.pem

Production private signing material is intentionally kept outside GitHub.

Standards and independent implementation

AML publishes a standards-oriented surface rather than requiring other runtimes to copy implementation internals:

• RFCs 0001–0011
• protocol discovery and registry
• JSON Schemas
• canonical serialization rules
• golden protocol vectors
• conformance fixtures
• layered compatibility levels: Core, Accountable, Federated, Verifiable, Governed
• explicit governance and standardization path
• independent second-runtime challenge

See ECOSYSTEM.md, STANDARDIZATION.md, and rfcs/README.md.

Quality gates

Every push runs repository-link checks, automated tests, deterministic receipt replay, balanced ALLOW/SUPPRESS fixture checks, public proof-surface guards, mainstream publication guards, verifier-contract drift checks, compiler verification, build-integrity verification, validation, semantic lint, explain/inspect checks, and benchmarks. A separate AML Conformance workflow verifies the public conformance surface.

Security

Security scope includes compiler/parser behavior, policy bypasses, receipt/signature integrity, wire replay, capability escalation, revocation, official trust-root verification, HTTP behavior, and browser-extension permission/privacy boundaries.

See SECURITY.md and SECURITY_THREAT_MODEL.md.

Evidence boundary

ĀML improves inspectability, policy control, reproducibility, interoperability, and cryptographic integrity.

It does not establish that current attention/restoration values objectively measure human cognition or wellbeing; it does not make an AI's declared intent truthful; its accessibility mechanisms do not replace WCAG conformance or assistive-technology testing; and cryptographic signatures do not prove moral correctness or institutional trustworthiness.

The question

When AI generates an interface, can the system prove what it meant, what authority it had, which policies governed it, and why the human saw the result?

Created by Daniel Jacob Read IV and stewarded by ĀRU Intelligence Inc.™.

ĀML™, ĀRU Meaning Language™, AI Interface Firewall™, View Meaning™, Meaning Gate™, EthicalRenderGate™, Meaning-Native Computing™, Proof-Carrying Interface™, and named ĀML compatibility marks are claimed marks. Registration status varies; do not use ® unless a specific mark is actually registered for the relevant goods/services.