# LGD Global Standards Alignment Matrix · v1.0 · 2026-09-06

> Purpose: anchor LGD on the existing network of global standards and regimes so the doctrine is verifiable, interoperable, and internationally credible. Alignment levels: **R** = explicit/structural mapping (the framework itself speaks the same language); **I** = informative mapping (relevant, to be deepened); **P** = planned engagement.
> Reviewer note: standard numbers are public and stable; revision/version dates were spot-checked at 2026-09-06 and must be re-checked against official issuers before formal academic reuse.

## A. Registry (LGD-I 有籍) — "who it is, who created it, who governs it"

| Standard / regime | Issuer | Alignment | What it gives LGD |
|---|---|---|---|
| ISO/IEC 42001:2023 AI management system (clause: AI system inventory & governance) | ISO/IEC | R | Management-level registry anchor |
| Algorithm registration/filing regimes (CN CAC algorithm filing; EU AI Act Art. 49 high-risk registration in EU database) | National/EU | R | Deployment registration exists in law — LGD supplies lifecycle continuity |
| EU AI Act Annex III (high-risk categories) | EU (Regulation 2024/1689) | R | Risk-tier → registry depth linkage |
| W3C DID / Verifiable Credentials | W3C | I | Machine-readable identity layer |
| FATF R.15 (VASP licensing/registration) | FATF | R | Entity registration for virtual-asset service providers |

## B. Evidence (LGD-II 有证) — "every consequential act leaves replayable evidence"

| Standard / regime | Issuer | Alignment | What it gives LGD |
|---|---|---|---|
| ISO 13485 (design/change records), ISO 14971 (risk file) | ISO | R | The original evidence-file paradigm |
| EU MDR 2017/745 & IVDR 2017/746 technical documentation | EU | R | Conformity evidence structure |
| FDA SaMD / TPLC framework, PCCP (predetermined change control plan) | US FDA | R | Change-accompanied-by-evidence in regulation |
| EU AI Act Art. 9-15 (risk mgmt, data governance, technical documentation, logging Art. 12, human oversight Art. 14) | EU | R | AI-side evidence duties now in law |
| NIST AI Risk Management Framework 1.0 + Generative AI Profile (Govern/Map/Measure/Manage) | NIST (US) | R | Voluntary but canonical evidence disciplines |
| eIDAS (EU 910/2014) / e-Evidence Reg. (EU) 2023/1543 | EU | R | Timestamp/signature/evidentiary chain standards |
| Electronic-evidence rules (CN SPC internet-court rules Art. 11(2); evidence rules Art. 93-94) | CN SPC | R | Hash/timestamp/blockchain/third-party custody recognized |

## C. Gates (LGD-III 有门禁) — "evolution passes assessment; humans authorize high stakes"

| Standard / regime | Issuer | Alignment | What it gives LGD |
|---|---|---|---|
| EU AI Act Art. 43 conformity assessment; Art. 26 deployer duties; Art. 27 FRIA | EU | R | Release + human-oversight gates in law |
| FDA PCCP & EU AI-DSF change-control thinking (AI/ML device changes) | FDA / EU | R | "Change = gate" precedent for software |
| ISO/IEC 42001 change/continuity controls + continual improvement | ISO/IEC | R | Management gate cycle |
| ISO/IEC 23894 AI risk management guidance | ISO/IEC | I | Gate scoring method input |
| Human-oversight gate theory (A³ in our doctrine family; EU Art. 14; CN 金发〔2026〕8号 high-risk AI approval by risk committee + human intervention) | — | R | "Machines request, humans authorize" echoed in multiple regimes |

## D. Lifecycle completeness & ethics (why "full-lifecycle" is the defensible claim)

| Instrument | Issuer | Use in LGD |
|---|---|---|
| Medical device full-lifecycle regulation (CN/FDA/EU) | CN / US / EU | Original template (see flagship paper §2) |
| UNESCO Recommendation on the Ethics of AI (2021-11) | UNESCO | Human-centricity baseline for the doctrine |
| WHO guidance on ethics & governance of AI for health (2021; large multimodal models 2024) | WHO | Health-domain authority anchor |
| OECD AI Principles | OECD | International-policy compatibility |
| Domain regimes mapped in `docs/` (FIN/LAW/GOV/CRYPT/AUT/DAT/UAS/ROB/IND/BIO/EDU) | various | Each domain paper cites its own verified mother-model |

## E. What this matrix buys us ("authority" operationalized)

1. **Verifiability**: every LGD claim can be checked against a named public standard/regime — the doctrine is auditable, not asserted.
2. **Interoperability**: LGD does not invent new law; it organizes existing duties into a lifecycle spine — low friction for adopters.
3. **Global coverage**: alignment spans CN, EU, US, ISO/IEC, FATF, UNESCO, WHO, OECD — no single-jurisdiction dependence.
4. **Gap-finding engine**: each empty cell in the matrix marks where "full-lifecycle continuity" is absent — that gap is our standing position.

## F. Maintenance rule

Re-verify standard versions against official issuers quarterly; update matrix and re-tag the repository; Zenodo concept DOI always points to the latest version.