Auditors: start at Trust Center
Public-safe rail

AI Management System (AIMS)

How MVG defines AI risk, assigns stop authority, and continuously improves — as receipts. Missing evidence never yields silent PASS (fail-closed).

Truth strip Reviewer routing v80.5

Single-source reviewer navigation. If any page text conflicts with a signed receipt, treat it as HOLD and follow the receipt.

/trust/reviewer-index/ Reviewer Index Key roles Site‑release latest
/.well-known/mvg-reviewer-index.json Machine index Signature
Canonical pointer /.well-known/mvg-prod.json (+ .asc) is the single signed production pointer. Determine state from signature validity (fail‑closed): missing ⇒ HOLD; invalid ⇒ FAIL; valid signature ⇒ proceed to verify referenced receipts offline.
Key roles site_release_verification_keyring_url = /trust/site-release/latest/pgp.asc · security_contact_encryption_key_url = /pgp.asc · security.txt
Verifier states PASS: all required proofs verify · HOLD: proof is missing, withheld, or unverifiable (fail‑closed) · FAIL: proof is present but invalid or inconsistent.
Open governance Open Trust Center Security review packet
AIMS descriptor (receipt)
Machine-readable AIMS surface (public-safe). Detached signature: /.well-known/mvg-aims.json.asc. Missing or placeholder ⇒ HOLD (fail‑closed by design).
HOLD
Checking publication…
sha256:— sha256:—
Open JSON Open .asc Open DSSE envelope Signing & audit runbook
Offline verify: gpg --verify .well-known/mvg-aims.json.asc .well-known/mvg-aims.json

What it is

An organization-level system for AI risk: policy → evidence → decision. The output is replayable, procurement-grade proof that a release was governed — not just asserted.

  • Defines “what must be proven” before a release can GO.
  • Standardizes HOLD triggers (uncertainty is treated as risk).
  • Emits signed descriptors and evidence pointers for auditors.

Who can stop a release

Stop authority, escalation, and incident posture are governed and published. Any missing or ambiguous evidence MUST produce HOLD by design.

/governance/ Open
/.well-known/mvg-governance.json Open

Continuous improvement loop

AIMS is operated as a closed loop aligned to common audit language (without claiming certification):

  • Govern: decision rights, escalation, and stop authority.
  • Map: intended use, affected stakeholders, and risk surfaces.
  • Measure: tests, controls, and replayable verification outputs.
  • Manage: remediation, change control, and release gates.

Public-safe boundary

This page and descriptor are public-safe: they explain the system and verification surfaces, while withholding sensitive internal details unless required for diligence.

  • Public: roles, escalation paths, verification surfaces, and receipts.
  • Withheld: internal personnel lists, private ticket contents, and confidential customer data.
  • Non-binding unless incorporated into a signed agreement.

Receipts

Discoverable

Auditors should treat signed descriptors as authoritative. Web pages are display-only.

/.well-known/mvg-aims.json Open
/.well-known/mvg-aims.json.asc Open

Verify (expected once GO‑LIVE): gpg --verify mvg-aims.json.asc mvg-aims.json

Expected outputs

Same language everywhere: verifiers output only one of three states.

  • PASS (GO): signature + evidence are present and valid.
  • HOLD: signature or required evidence is intentionally missing (fail‑closed).
  • FAIL: a required signature/evidence is present but invalid or inconsistent.
Receipt Finder Verify offline