Skip to content

ĀML Core 1.3.0 — Trust, Difference, Accessibility, and Runtime Accountability

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 08 Sep 03:50
v1.3.0
7c5f2ad

ĀML Core v1.3.0 — Trust, Difference, Accessibility, and Runtime Accountability

ĀML v1.3.0 extends Accountable AI Execution with six new system-level capabilities:

  • signed policy packs
  • semantic diffs
  • policy diffs
  • hash-chained runtime audit streams
  • executable accessibility policies
  • cumulative attention accounting

Signed policy packs

Policy packs are data-only compositions of installed policy IDs. They can be canonically hashed, signed with Ed25519, and independently verified. The design intentionally does not allow a signed policy pack to smuggle executable JavaScript.

Semantic diffs

aml semantic-diff compares Abstract Meaning Trees and reports added, removed, changed, and unchanged meaning-bearing nodes, separating structural, property, and meaning-metadata changes.

Policy diffs

aml policy-diff runs identical source/context through two policies or policy profiles and reports the exact render decisions whose outcome or rationale changed.

Runtime audit streams

ĀML now provides append-only SHA-256 hash-chained audit streams. Each runtime event commits to the previous event hash, making mutation detectable.

Accountable execution receipts can bind the audit stream and its verification result.

Cumulative attention accounting

The new Attention Ledger tracks attention across a session rather than treating every node independently. Once earlier rendered components consume the budget, later components can be suppressed by cumulative enforcement.

Accountable execution regenerates final HTML from the cumulative-enforced decision set.

Accessibility policies

New policies:

  • reduced_motion_v1
  • contrast_safety_v1
  • cognitive_load_guard_v1

New profiles:

  • accessibility_first
  • human_first

These mechanisms are an executable policy layer and do not claim to replace WCAG conformance or assistive-technology testing.

New CLI surface

aml semantic-diff
aml policy-diff
aml sign-policy-pack
aml verify-policy-pack
aml verify-audit
aml attention-account

These join the existing compiler, accountable execution, receipt signing, conformance, integrity, diagnostics, language-server, and browser tooling.

Repository reliability

v1.3 also adds a local Markdown link-integrity gate to CI. Broken relative repository links now fail the build instead of silently producing public GitHub 404s.

Compatibility paths were restored for:

  • REPLICATION.md
  • CONFORMANCE.json
  • docs/API.md

Verify

git clone https://github.com/aruintelligence/aml-core.git
cd aml-core
npm install --ignore-scripts
npm run check:links
npm test
node bin/aml.js version
node bin/aml.js policies
node bin/aml.js profiles

Browser playground:

https://aruintelligence.github.io/aml-core/playground.html

Live lab:

https://aruintelligence.github.io/aml-core/

Evidence boundary

ĀML v1.3 improves policy transparency, integrity, runtime accountability, accessibility-aware decision inputs, and cumulative accounting. It does not establish that any current policy is universally ethical, that attention/restoration scores are validated human measurements, or that a signature proves an issuer is trustworthy.

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

Code is MIT licensed. Claimed marks remain separate from the code license.

Assets 2
Loading