Most Microsoft 365 tenants aren't breached because of some exotic zero-day. They're breached because a handful of default settings were never changed, a legacy protocol was never turned off, or a global admin account sat there without MFA for two years. The same Microsoft 365 security misconfigurations show up again and again — across MSP-managed tenants and internal IT environments alike.
This post walks through the patterns we check for in every M365 / Entra ID audit, why each one matters, and how to confirm whether your own tenant is affected. None of these require a rip-and-replace project. Most are a settings change and a short validation window.
1. MFA gaps on the accounts that matter most
The single most common finding is not "MFA is off everywhere." It's MFA enforced inconsistently — enabled for most users but missing on the exact accounts an attacker wants:
- Break-glass / emergency access accounts (sometimes legitimately excluded, often just forgotten)
- Service accounts and shared mailboxes repurposed as interactive logins
- Recently created admin accounts that slipped past the enrollment policy
- Guest accounts with elevated access
Security defaults or a Conditional Access policy can cover the baseline, but the gaps live in the exceptions. The fix is to enumerate every account with any administrative role and confirm each one is either MFA-enforced or a deliberately documented exclusion. "We turned on MFA" is not the same as "every privileged identity is covered."
2. Legacy authentication still enabled
Legacy auth protocols — IMAP, POP, SMTP AUTH, and older Exchange/Office clients — can't present a second factor. That means they completely bypass most Conditional Access MFA requirements. A tenant can have flawless MFA policies and still be wide open if legacy auth is reachable.
This is one of the highest-severity findings precisely because it's invisible from the admin's point of view: the MFA policy looks enforced. Password-spray attacks specifically target legacy endpoints for this reason. The remediation is to block legacy authentication via Conditional Access (and confirm nothing critical actually depends on it first — printers, scanners, and old line-of-business apps are the usual casualties).
3. Conditional Access policies that look complete but aren't
Conditional Access is powerful and easy to misconfigure. The recurring patterns we flag:
- No baseline "require MFA for all users" policy, only targeted ones that leave holes
- Report-only policies that were created, tested, and never moved to "On"
- Exclusion groups that quietly grew to include dozens of users
- No sign-in or sign-up risk policies, so Entra ID's own risk signals go unused
- Missing device compliance or managed-device requirements for admin actions
A tenant can have a dozen Conditional Access policies and still have a straight path to a privileged mailbox. The value of an audit here is tracing the effective access — what actually gets enforced when a real user signs in — not just counting how many policies exist.
4. Over-privileged and standing admin access
Role sprawl is nearly universal. Common findings:
- More Global Administrators than the tenant needs (Microsoft's guidance is to keep this small — single digits for most orgs)
- Permanent (standing) admin assignments where Privileged Identity Management (PIM) just-in-time elevation would be far safer
- Admin roles granted to day-to-day user accounts instead of dedicated admin identities
- Stale admin accounts for people who changed roles or left
Every standing Global Admin is a full-tenant compromise waiting for one phished password. The fix is a least-privilege pass: reduce Global Admins, move to scoped roles where possible, and adopt PIM so elevated access is time-boxed and logged.
5. Sharing and collaboration defaults left wide open
SharePoint and OneDrive ship with permissive sharing defaults that many tenants never tighten:
- "Anyone" (anonymous) sharing links enabled org-wide
- Links that never expire
- External sharing allowed to any domain rather than an allowlist
- Guest access with no access-review cadence
These rarely cause an obvious incident, so they're easy to ignore — until a sensitive document is indexed or forwarded outside the org. The remediation is usually to set the most permissive link type to "specific people," add expiration, and restrict external sharing to known partner domains.
6. Mail flow and anti-phishing gaps
Email remains the primary initial-access vector, and the configuration gaps are consistent:
- No DMARC record, or DMARC in
p=nonewith no plan to move to quarantine/reject - SPF and DKIM misaligned or missing on sending domains
- Auto-forwarding to external addresses left enabled (a classic exfiltration and BEC path)
- No or weak anti-phishing policies in Defender for Office 365
DMARC at p=none is a particularly common "we did the project" illusion — the record exists but enforces nothing. It only reports, so a spoofed message still lands in an inbox exactly as before. Moving to p=quarantine or p=reject is the step that actually closes the gap, and it only takes confidence that the alignment reports show legitimate mail isn't being caught too.
The short version
None of these are obscure. They're the same handful of settings and policies showing up again and again across different tenants, because the defaults were never revisited and nobody is checking on a schedule. Finding them is mostly a matter of knowing where to look — Conditional Access, role assignments, sharing defaults, and mail flow — and confirming what's actually enforced, not just what's configured.
Run an audit to check your tenant against these patterns and the frameworks they map to, in minutes.