Compliance blueprints: framework minimums for every laptop
A blueprint is the set of settings Shadow MDM enforces on your devices. You do not write it rule by rule: you pick the frameworks your organisation follows, and Shadow compiles the blueprint from each framework’s minimum, plus any stricter Proposals you approved, minus exceptions you accepted. Every approved change makes a new, signed blueprint version.
Framework minimums
Section titled “Framework minimums”Shadow groups rules into two building blocks, then adds framework-specific requirements.
- Endpoint baseline (ISO 27001, SOC 2, GDPR; also part of PCI DSS and HIPAA): disk encryption, screen lock after 15 minutes, host firewall, automatic OS updates, OS supported and patched (critical patches within 14 days, at most one major version behind), anti-malware with current definitions, no day-to-day admin account, local passwords of at least 8 characters, removable storage reported, remote-wipe readiness, software inventory, CIS Level 1 hardening.
- AI blueprint (ISO 42001, EU AI Act): Shadow agent installed and tamper-protected, AI traffic through Shadow, personal data masked before AI, only approved AI tools, extensions and MCP servers, local model runtimes reported, software inventory.
| Rule | ISO 27001 | SOC 2 | GDPR | PCI DSS v4.0.1 | HIPAA | ISO 42001 | EU AI Act |
|---|---|---|---|---|---|---|---|
| Disk encryption, recovery key escrowed | ✓ | ✓ | ✓ | ✓ | ✓ | ||
| Screen lock after idle | 15 min | 15 min | 15 min | 15 min | 10 min | ||
| Host firewall on | ✓ | ✓ | ✓ | ✓ | ✓ | ||
| Automatic OS updates | ✓ | ✓ | ✓ | ✓ | ✓ | ||
| Critical patches within | 14 days | 14 days | 14 days | 14 days | 14 days | ||
| Anti-malware definitions at most | 3 days old | 3 days old | 3 days old | 1 day old | 3 days old | ||
| No daily admin account | ✓ | ✓ | ✓ | ✓ | ✓ | ||
| Local password length | 8 | 8 | 8 | 12 | 8 | ||
| Removable storage | report | report | report | report | report | ||
| Remote lock and wipe ready | ✓ | ✓ | ✓ | ✓ | ✓ | ||
| Software inventory | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| CIS benchmark hardening | Level 1 | Level 1 | Level 1 | Level 1 | Level 1 | ||
| Shadow agent tamper-protected | ✓ | ✓ | |||||
| AI traffic through Shadow | ✓ | ✓ | |||||
| Personal data masked before AI | standard | standard | |||||
| Approved AI tools only | ✓ | ✓ | |||||
| Local model runtimes | report | report |
PCI DSS also allows critical patches within 30 days; because the merge keeps the strictest value, the endpoint baseline’s 14 days applies when PCI DSS is combined with it.
Strictest wins
Section titled “Strictest wins”When you follow more than one framework, Shadow merges them per rule and keeps the strictest value: the shortest idle time and patch window, the longest password, block over report for removable storage, strict over standard masking, approve over report for local model runtimes. A workspace on ISO 27001, HIPAA and PCI DSS gets a 10-minute screen lock, 12-character passwords and daily anti-malware definitions.
ISO 42001: the AI blueprint
Section titled “ISO 42001: the AI blueprint”When you select ISO/IEC 42001, the AI blueprint is always on. It is where Shadow MDM differs most from a traditional MDM: the same agent that masks data going to AI enforces the AI controls, so you can show that AI traffic goes through Shadow, that personal data is masked, and that only approved AI tools and MCP servers are in use. See ISO 42001 for AI on laptops.
CIS Level 1 vs Level 2
Section titled “CIS Level 1 vs Level 2”The CIS Benchmarks define two hardening profiles.
| CIS Level 1 (default) | CIS Level 2 (“Strict”, optional) | |
|---|---|---|
| Goal | Practical hardening with little impact on daily work | Defence in depth for high-security environments |
| macOS examples | Gatekeeper on, System Integrity Protection on, guest account off, sharing services off | Level 1, plus Bluetooth sharing and AirDrop off, and more |
| Windows examples | SmartScreen on, guest account off, SMBv1 off | Level 1, plus further restrictions |
| Proposed as a minimum | Yes, for ISO 27001, SOC 2, GDPR, PCI DSS and HIPAA | Never; an admin chooses it |
| Disruption | Low | Can break everyday workflows |
Shadow proposes Level 1 because the frameworks ask for a hardening baseline, not for the strictest profile. If your risk assessment calls for Level 2, choose the Strict option; Shadow then writes Proposals for the extra settings, rolled out like any other change.
Beyond the minimum
Section titled “Beyond the minimum”On top of the minimum, Shadow proposes tighter settings where the data says so. Laptops whose AI traffic is the most sensitive can get, for example, a 5-minute screen lock, blocked removable storage and strict masking. These are Proposals too: you approve them, scoped to those devices only.
Platforms
Section titled “Platforms”Each rule maps to native settings on macOS, Windows and Linux. Two limits today:
- macOS remote lock and wipe need the Apple MDM protocol; Shadow’s Apple MDM certificate is pending. Until then, the remote-wipe check on Macs reports as not ready and lock uses the Shadow agent’s lock screen.
- Linux disk encryption (LUKS) is report-only. Some other Linux checks, such as password policy and anti-malware, are report-only as well.
Next: the Enforcement Map and the rule-to-control mapping.
Frequently asked questions
Can a model lower a framework minimum?
No. The minimum comes from the frameworks your workspace follows and is fixed. Models only decide how, when and for which devices a setting lands, and can propose stricter settings, never weaker ones.
Why is CIS Level 1 the default and not Level 2?
CIS Level 1 is the practical hardening baseline that rarely gets in people's way. Level 2 adds settings, such as turning off Bluetooth sharing and AirDrop, that can break everyday work. Shadow never proposes Level 2 as a minimum; an admin can choose it as the Strict option.
What if we need an exception for one laptop?
Decline the Proposal for that device with a reason. Shadow records it as an exception (a risk acceptance) with who decided and when, removes it from that device's blueprint and shows it on the Enforcement Map.
Do blueprints cover every control in ISO 27001 or SOC 2?
No. A device MDM covers endpoint controls only. Controls it cannot cover are marked out of scope on the Enforcement Map, so coverage is stated honestly.