Alchemy Playbooks · In development

AI Compliance Mapping:
one control, mapped across every framework.

An AI governance crosswalk for the ISO/IEC 42001, the EU AI Act and NIST AI RMF: AI compliance mapping with the evidence an auditor would ask for. Being built now. Launch date to be confirmed.

3
frameworks in
version 1
ISO 42001 · EU AI Act · NIST AI RMF
38
Annex A controls
in ISO/IEC 42001
ISO/IEC 42001:2023
Dec 2027
agreed extended deadline for
Annex III high-risk AI systems
Digital Omnibus on AI, May 2026, pending formal adoption
TBC
launch date, to be
confirmed after beta
Updated here as it firms up

Three rulebooks.
One underlying control.

Organisations using AI face several overlapping rulebooks at once. The EU AI Act is law. NIST AI RMF is voluntary guidance. ISO/IEC 42001 is a certifiable management system standard. Each has its own structure and vocabulary, so the same underlying control tends to get documented three times.

AI compliance mapping lines them up against each other. You can see where one control answers several frameworks, where it only partly does, and where a framework asks for something the others do not.

It is a companion to MASTER-AI ™: where MASTER-AI structures the governance function, the compliance mapping shows how each control lines up with the standards and laws that apply, with every control organised under the six MASTER-AI domains. It supports compliance work and is indicative guidance, not legal advice.

The AI Compliance Mapping CrosswalkThe minimum number of controls to maximise regulatory coverage

Three things that make it
more than a lookup table.

1
Beyond the clause reference

Auditor Evidence

For each control, the exact audit artefacts an auditor would expect to see. A bare mapping of identifiers is easy to find; knowing what to put in front of an auditor is the part most mappings leave out.

2
Common baseline · Jurisdiction deltas

Baseline Plus Deltas

Rather than treat each framework separately, the aim is the fewest possible unified controls, anchored on the most demanding regime, then the extra requirements each other regime adds on top. Whether that holds up is exactly what the beta is there to test.

3
Plain language · Dated · Reviewed

Own Wording, Versioned

Descriptions are written in plain language and keyed to published control identifiers, so no standard text is reproduced. Each release is dated and reviewed, with a change log, because a crosswalk that goes stale is worse than none.

Version 1 first.
More jurisdictions to follow.

Version 1 covers the three frameworks that most organisations meet first. Later versions are planned to add further jurisdictions, with exact instruments to be confirmed during research.

Version 1 · Standard

ISO/IEC 42001:2023

Management system clauses and Annex A controls, described in original wording.

Version 1 · Law

EU AI Act

The high-risk obligations first, as the structural spine for the common baseline.

Version 1 · Guidance

NIST AI RMF

The Govern, Map, Measure and Manage functions, with their categories and subcategories, mapped to the baseline.

Later version · Candidate

Singapore

AI governance framework and related guidance. Instruments to be confirmed.

Later version · Candidate

California

State AI laws relevant to organisations operating in or selling into California. Instruments to be confirmed.

Later versions

Further jurisdictions

Each version number adds more frameworks, driven by what readers ask for.

Where it stands,
updated as it moves.

01 · Concept and Scope Done

Scope, methodology, and hosting finalised.

The Crosswalk is built and permanently hosted on GitHub, with a structured spreadsheet and an MCP as outputs. Two differentiators:

  • Coherence: Grounded in MASTER-AI ™.
  • Practicality: Evidence-first, specifying the exact audit artefacts needed to demonstrate compliance.
02 · Beta and Consultation Next

Test, prove and perfect through collaboration

A small beta community, ideally including: (a) a first-party enterprise auditor, (b) a second-party AI governance consultancy, (c) a third-party AI auditor, to test the Excel artefact, and (d) an AI software vendor, to test the MCP endpoint. Please do reach out to join.

03 · Full Version 1 Launch Planned

Publish the dataset openly

An Excel and CSV artefact for people and an MCP remote server for machines, mapped, reviewed and released under a clear licence. A major release at least quarterly, with minor releases as regulations change.

04 · Future Extension Under consideration

Target the golden thread

Help practitioners “build once and run anywhere”: one AI governance programme that meets the requirements of every jurisdiction it operates in. Likely extensions: US states (for example California) and country-level models (for example Singapore).

David Viney, Fractional CIO and AI Transformation Director
“Our goal is to build an AI Compliance Mapping that achieves maximum regulatory coverage using the fewest possible unified controls.”
Working Spaces

Collaborate on building the Body of Knowledge

The controls, evidence and releases are built in the open on GitHub. Pick the space that suits you.

AlchemyMCP Server
AI Compliance Mapping
MCP Server

Refine the beta MCP endpoint and validate integrations across target LLMs and software tools.

Join discussion
AlchemyPlaybook
AI Compliance Mapping
Crosswalk

Challenge the mapping. Tell us where an auditor would disagree, or which evidence is missing.

Raise an issue

Watch it get built.
Tell me what would help.

Progress and versions are tracked in the open on GitHub, and summarised on this page. If you work on AI assurance, audit or governance and would find this useful, want to suggest ideas, would like to join the beta, or want to talk about using it, I would like to hear from you.

In development · Open resource
Follow the build

On GitHub, choose Watch, then Custom, then Releases to hear about new versions, or star the repo to show support.

The compliance mapping will be indicative guidance to support planning and audit preparation. It is not legal advice and does not replace reading the source texts.