# ISO 42001 Model Development

> Draft model development documentation for ISO/IEC 42001:2023 (architecture, training, testing, limitations).



Tags: ISO 42001, Model Development, Documentation, AI


## Example Prompts

- Document our model architecture and training data
- Draft testing results and limitations for audit
- Update model development documentation

URL: https://rakenne.app/skills/iso42001-model-development/index.md

Try this skill: https://rakenne.app/a/?skill=iso42001-model-development



## Overview

Draft model development documentation for ISO/IEC 42001:2023 (architecture, training, testing, limitations).

## Getting Started

Add your source materials to the project workspace, then activate the *ISO 42001 Model Development* skill. The agent will guide you through the process step by step.

## Example conversation

**You:** We need to document our model development for the recommendation engine—architecture, training data, testing, limitations, and security—for the auditor.

**Agent:** I’ll draft the model development documentation using the skill reference. I’ll describe architecture and design decisions, training data sources and preparation, testing results (functional, bias, robustness), limitations and known risks, transparency/explainability methods, and information security safeguards. I’ll run **model_development_check** on the document and fix any missing elements, then link to your data governance and risk register.

## Output excerpt

### Model development documentation excerpt

**Architecture** — [e.g. two-tower retrieval + ranking model]. Design decisions: [e.g. offline training, online inference]; [e.g. feature set and constraints]. Assumptions: [e.g. stationarity of behaviour within retrain window].

**Training data** — Sources: [e.g. engagement events, segment attributes]; period [X]. Preparation: [e.g. filtering, aggregation]; see data governance for provenance and quality. No special-category data.

**Testing** — Functional: [e.g. accuracy, coverage]. Bias: [e.g. demographic parity, segment fairness]. Robustness: [e.g. stress tests]. Results summarized in [e.g. model card or test report]; limitations documented below.

**Limitations** — Cold-start; possible bias in long-tail segments; not validated for [e.g. eligibility or safety-critical] use. Mitigations: monitoring, retrain cycle, human oversight.

**Transparency** — [e.g. Model card]; in-app disclosure; [e.g. feature contribution or surrogate explainability] where requested.

**Security** — [e.g. model and data access control; pipeline in secure env; no model export to untrusted clients.]

## Extension and validation

The skill includes **model_development_check**, which validates model development documentation for required elements: model architecture and design decisions; training data sources and provenance; testing results (functional, bias, robustness); limitations and known risks; transparency/explainability methods; information security safeguards. Run it after drafting and address any missing elements.


---

Back to [Skill Library](https://rakenne.app/skills/index.md)