Security & data protection
Your Playbook Is Your Moat. We Treat It That Way.
MagiLearn OS holds your operating knowledge — SOPs, standards, training records. The protections below are architectural decisions, not bolted-on promises: each one is enforced in code and covered by automated tests.
Six Practices, Built Into The Architecture
Tenant Isolation On Every Table
Every business record carries a tenant id and every query filters by it — workspace data is never mixed. Custom domains and white-label branding never weaken isolation.
Permissions Enforced At The Data Layer
Role permissions are checked in the API and data layer on every request — hiding a button is never the security boundary. Course visibility is enforced in the library, global search, AI retrieval and recommendations alike.
AI That Cannot Leak
The assistant retrieves only within the asker's permissions — how closed retrieval works lives on the AI page.
How closed retrieval works →Audit Trail For Sensitive Actions
Invitations, role changes, deactivations, policy acknowledgements and other sensitive actions are recorded with actor, timestamp and before/after state.
PDPA-Aligned Data Rights
Per-person data export and erasure are built in. Deactivating a member revokes all access immediately; learning records follow soft-deletion with retention rules.
Least-Privilege Operations
Secrets live in environment configuration, never in code. Uploads are treated as untrusted input, and AI-generated content always requires human approval before it reaches learners.
Where The Isolation Actually Happens
Not in the interface — in the data layer. Every query carries the workspace and the asker's permission keys, so the wrong content can't be fetched, not merely can't be seen.
Your Workspace (HQ)
Internal courses, SOPs, people
tenant_id
Partner Portal
Only what you assign to it
tenant_id
Another Company
A different world entirely
tenant_id
One data layer — every query filtered by workspace and permission key
Enforced in the course library · global search · AI retrieval · recommendations — and covered by automated tests
Your Data Stays Yours
Data rights are product features, not support tickets.
Export Anytime
People, learning records and reports export to CSV on every plan. Your operating knowledge is yours — leaving must never mean losing it.
Pause, Don't Lose
If a subscription lapses, the workspace pauses instead of vanishing. Nothing is deleted for 60 days, so a billing hiccup never destroys training history.
Erase On Request
Per-person erasure removes personal data while preserving the anonymous compliance evidence your business must keep.
Security Questions, Answered Directly
Who can see our workspace content?
Only members of your workspace, within the permissions you set. Course visibility scopes are enforced at the data layer — in the library, global search, AI retrieval and recommendations — and external partner portals are isolated from HQ content by design.
What happens to our data if we leave?
Your data exports with you: people, learning records and content are available as exports at any time, on any plan. If a subscription ends, the workspace pauses and nothing is deleted for 60 days.
How does MagiLearn OS handle personal data under PDPA?
Per-person data export and erasure are built into the product. Deactivating a member revokes all access immediately, and learning records follow soft-deletion with retention rules rather than silent hard deletes.
Found a vulnerability, or running a due-diligence review? Reach us via the contact page — security reports get a response within one business day, and we credit responsible disclosure.