Safety Nexus / Executive brief
GB Standards AI Reviewer
An AI-assisted standards review concept for finding, comparing and explaining requirements while accountable technical professionals retain the final judgement.
Public proof case / 03
Use this case to assess high-trust AI design: retrieval is useful only when jurisdiction, edition, scope, citations and accountable review remain visible.
Evidence atoms
Problem
Requirements cross contexts
Teams need to retrieve and compare requirements without losing scope, definitions or project facts.
Alvin's role
High-trust workflow builder
The public portfolio identifies Alvin Liao as the practitioner behind this standards-review direction.
Workflow
Retrieve, compare, review
Scope the question, retrieve sources, draft comparison notes, then route to technical review.
Controls
Citations and grounding
Source grounding and reviewer discipline are described; version, licensing and audit details are not public.
Safe boundary
Review support only
Support search and explanation; do not decide compliance, approve design or replace engineering judgement.
Publicly verifiable
Direction, not certification
No accuracy rate, regulator approval, certification or independent audit result is claimed.
Review workflow
Define applicability
Identify jurisdiction, edition, project context and the consequence of error.
Find candidates
Narrow the question toward relevant standards material and clauses.
Separate layers
Keep source facts, semantic differences and interpretation visibly distinct.
Return judgement
Check citations and context before a qualified professional decides.
Controls and boundaries
Governance controls
Inspect the evidence trail
- Keep source and clause references visible.
- Verify edition, scope, definitions and project conditions.
- Protect licensed content and retain an auditable review handoff.
Safe boundary
Prepare, do not certify
Suitable support includes retrieval, comparison and draft notes. This proof does not establish compliance, legal advice, regulator acceptance, design approval, incident findings or shutdown authority.
Public verification and lessons
What can be cited
The public portfolio describes retrieval and semantic-review support across GB, TB and HKRS requirements and links the direction to standards-aware AI implementation. The live configuration and application link are not presented here.
Lesson
Applicability beats retrieval speed. The right clause is not enough if the edition, scope or project condition is wrong.
Related official sources
Official GB standards portal.
Official HKRS reference.
AI-risk governance context.
Official references provide context; they do not certify this reviewer or establish compliance.