Learn › Power BI & Fabric › 3 · Governance & security

The 170+ tenant settings and governing them

Fabric's admin portal holds 170+ tenant settings, and a handful of them are the difference between governed analytics and a data-leak waiting to happen. The ones that matter, how delegation works, and why they drift.

The settings that actually matter (of the 170+)

Most are benign feature toggles; a small set are security-critical:

Setting Why it's dangerous Sane posture
Publish to web Makes a report PUBLICLY readable on the internet, no auth — the classic accidental data leak Disabled, or tightly allowlisted
Export data / Export to Excel/CSV Data walks out of the governed surface Scoped to groups that need it
External sharing / guest access to content Data reaches outside the tenant Off or allowlisted per the org's stance
Allow service principals / developer APIs Programmatic access to the whole estate Scoped to specific security groups, audited
Information protection / sensitivity labels Whether labels apply to Power BI content On — the Purview tie-in

How delegation works

Most settings support Enabled for the entire org / Disabled / Enabled for specific security groups (with exceptions). The governed pattern: disabled by default, enabled for a named group with a documented reason — the same least-privilege shape as every other workload. The trap is 'Enabled for the entire org' left on a dangerous toggle (publish-to-web especially) because nobody reviewed the default.

Why they drift — and why that's serv365's whole point

170+ settings, multiple admins, feature launches flipping defaults, and a 'temporary' enablement that outlives its reason — this surface drifts constantly. serv365 scans all 171 Fabric tenant settings nightly, canonicalises and diffs them: publish-to-web flipping on, an export control loosening, a new service-principal grant — each is a dated, attributed change, and the dangerous ones should page someone. A point-in-time 'we configured it right' is worth far less than 'nothing changed that we didn't intend'.

What to watch (proofs)

  • The dangerous subset as alerts: publish-to-web, export, external sharing, service-principal access — their current state (nightly scan) with priority alerting on change; these are the settings attackers and mistakes flip.
  • Delegation reality: for each critical setting, 'entire org' vs 'specific groups' — the org-wide-on-a-dangerous-toggle audit.
  • Publish-to-web exposure: the admin portal's publish-to-web report lists every publicly-exposed report — run it; the findings are often alarming.
  • Drift attribution: the Fabric activity log names who changed a tenant setting — an unexpected actor on a security toggle is an incident.

Discussion

No messages yet — start the thread.

Sign in with your email to join the discussion — we send a one-time link, no password.