ISO 42001 for AI on laptops: govern the AI tools people actually use
ISO/IEC 42001 asks an organisation to know which AI systems it uses, what data goes into them and who decided they were acceptable. On laptops that is hard: AI is in browsers, coding agents, IDE extensions, MCP servers and local model runtimes. Shadow already sits in the path of that traffic for AI privacy, so the Shadow MDM AI blueprint can enforce and evidence these controls from the same agent.
The AI blueprint
Section titled “The AI blueprint”When you select ISO 42001 (or the EU AI Act) in Shadow Hub, the AI blueprint is always on:
| Rule | What it checks or enforces | ISO 42001 | Also supports |
|---|---|---|---|
| Shadow agent protected | Shadow is installed, running and tamper-protected | A.9.2 | ISO 27001 A.8.16, EU AI Act Art. 26 |
| AI traffic through Shadow | Requests to AI services go through Shadow | A.9.2, A.10.3 | ISO 27001 A.8.12, GDPR Art. 32 |
| Personal data masked | Personal data is replaced before it reaches AI (standard or strict) | A.7.4 | ISO 27001 A.8.11, GDPR Art. 25 |
| Approved AI tools only | Only approved AI tools, extensions and MCP servers | A.4.4, A.9.4, A.10.3 | EU AI Act Art. 26 |
| Local model runtimes | Ollama, LM Studio and similar are reported, or need approval | A.4.5, A.9.4 | ISO 27001 A.8.19 |
| Software inventory | Installed software is inventoried | A.4.4 | ISO 27001 A.5.9 |
See the rule-to-control mapping for every rule.
Proactive AI governance
Section titled “Proactive AI governance”- New AI tools. When an AI tool without a policy shows up on two or more devices, Shadow writes a Proposal to approve or block it, with who uses it and how often.
- Sensitive users. Laptops that send the most sensitive data to AI get Proposals for stricter masking and settings, scoped to those devices.
- Local models. Local model runtimes are reported by default; you can require approval instead.
Every change is a Proposal an admin approves in Shadow Hub or Meridian, rolled out to a pilot group first with one-click rollback.
Evidence
Section titled “Evidence”The Enforcement Map shows ISO 42001 → control → rule → devices, with approvals, exceptions and history, and gives your auditor a read-only role and signed exports. The same results feed compliance platforms through the Evidence API as control SHD-13, AI tools on endpoints are governed.
With ISO 27001
Section titled “With ISO 27001”Most teams pursuing ISO 42001 also hold or are working towards ISO 27001. Select both and Shadow merges the endpoint baseline and the AI blueprint, strictest setting wins. See Pass ISO 27001 without an MDM.
Next steps
Section titled “Next steps”Frequently asked questions
What is ISO/IEC 42001?
The international standard for an AI management system (AIMS), published in 2023. Like ISO 27001 it is certifiable and built around policies, risk assessment and Annex A controls, here focused on how an organisation develops, provides or uses AI systems.
Does ISO 42001 apply if we only use AI tools and don't build them?
It can. The standard covers organisations that use AI systems as well as those that build them. Many teams start with the AI they use every day: chat assistants, coding agents, browser extensions and local models on laptops.
Does enforcing the AI blueprint cost extra?
No. Posture and AI inventory reporting, and enforcing the AI blueprint through approved Proposals, are part of Shadow MDM, which is included in Shadow Hub at $3 per user per month.
Does this make us ISO 42001 certified?
No. Certification covers your whole AI management system. Shadow supports the controls about AI resources, data and use on laptops, and provides evidence for them.