Conditional Access Mfa

Conditional Access & MFA

Conditional Access (CA) is the Zero Trust policy engine in Entra. It evaluates signals - user /group, device state, location, application, sign-in risk - and enforces grant, block, or grant-with-controls (MFA, compliant device, session controls). MFA without CA is rare; the two are designed together.

When to use

Enforcing adaptive, identity-centric access controls across Microsoft 365 and any Entra- integrated app. Use this skill for the baseline policy set and any new policy rollout.

Do not use this skill for risk detection (entra-id-protection), just-in-time admin elevation (azure-pim), or app-side OAuth permissions (entra-id).

The baseline 6-policy set

Every Entra tenant in 2026 should have these six policies as the minimum, in this order.

| # | Policy | Scope | Grant control | |---|---|---|---| | 1 | Require MFA for admins | All Entra admin roles | MFA + phishing-resistant strength | | 2 | Require MFA for all users | All users (excl. break-glass) | MFA | | 3 | Block legacy authentication | All users | Block | | 4 | Require compliant device for M365 | All users, M365 apps | Compliant OR hybrid-joined | | 5 | Risk-based CA - sign-in risk High | All users | MFA (requires EID P2) | | 6 | Risk-based CA - user risk High | All users | Secure password change (requires EID P2) |

> Rule of thumb: if a tenant doesn't have all 6, it is below the 2026 Microsoft baseline. > Policies 5 and 6 require Entra ID P2; substitute with risk-based authentication strength > on E3 if needed.

Approach

1. Inventory before authoring — Export current sign-in logs to understand baseline:

per-app MFA usage, legacy auth volume, device compliance, named locations in use. Don't author policies on assumed traffic patterns. Verify: Sign-ins workbook shows per-app MFA % and legacy auth count.

1. Use authentication strengths, not just "Require MFA" — Authentication strengths let

you require phishing-resistant methods (FIDO2 passkey, Windows Hello for Business, certificate-based) for high-value scopes. Plain "Require MFA" accepts SMS - still phishable. Verify: admin policy grant control uses the built-in Phishing-resistant MFA authentication strength, not generic MFA.

1. Always report-only first — New policies in report-only mode for at least 7 days.

Use What If to model specific user + app scenarios. Promote to On only after the report-only results show no surprises. Verify: policy report-only impact > 0 sign-ins evaluated (otherwise policy isn't matching anything - bad scope).

1. Exclude break-glass on every policy — Two cloud-only emergency accounts excluded from

every policy, monitored on every sign-in. Document the exclusion list and review monthly. Verify: Conditional Access Insights workbook shows break-glass exclusion applied to all enforcing policies.

1. Session controls where needed — Sign-in frequency for sensitive apps, persistent

browser block on shared devices, app-enforced restrictions for Exchange/SharePoint, Conditional Access App Control for in-session monitoring.

1. Policy register — Maintain a register (CSV or Sentinel watchlist) with policy intent,

owner, exclusions, last review. CA policies drift without this.

Guardrails

• Never enforce a new policy on All users without report-only first. Most production CA

incidents come from this single mistake.

• Two break-glass accounts excluded from every policy, monitored on every sign-in.

Without this, a CA misconfiguration can lock out the tenant.

• Block legacy auth, then monitor sign-ins for 30 days. Some apps still use it -

surface them and remediate before complaints arrive.

• Phishing-resistant strength for admins, not generic MFA. SMS-based MFA for Global Admin

is the 2026 equivalent of a password sticky note.

• Don't stack 30+ policies. Each evaluation costs latency. Consolidate where conditions

overlap; aim for < 20 active policies.

• Test changes with What If, not with a colleague's account. What If tests against the

policy engine without a real sign-in.

Common anti-patterns

• "New policy direct to On for All users" - Lockout. Report-only 7 days minimum.
• "Require MFA = phishing-resistant" - It accepts SMS. Use authentication strengths.
• "One break-glass account in a password manager" - Single point of failure. Two

hardware-key accounts, separated custody.

• "No exclusions because exclusions are insecure" - Without break-glass exclusions, you

can't recover from a CA misconfiguration.

• "Block all of Russia / China at named location" - Attackers use proxies in your

allowed countries. Use risk signals + device compliance instead.

• "CA policies in production with no register" - 6 months later no one knows why policy

17 exists.

Example prompts

• Author the baseline 6-policy Conditional Access set for a new Entra tenant.
• Require phishing-resistant MFA (FIDO2/WHfB) for all Entra admin roles.
• Block legacy authentication and report on which apps still use it.
• Roll out a new "require compliant device for Exchange" policy safely with report-only.
• Why are users getting MFA-prompted constantly - tune sign-in frequency.
• Exclude our break-glass accounts from every CA policy and alert on their sign-ins.

Microsoft Learn

• Conditional Access overview: https://learn.microsoft.com/entra/identity/conditional-access/overview
• Authentication strengths: https://learn.microsoft.com/entra/identity/authentication/concept-authentication-strengths
• Report-only mode: https://learn.microsoft.com/entra/identity/conditional-access/concept-conditional-access-report-only
• What If tool: https://learn.microsoft.com/entra/identity/conditional-access/what-if-tool
• Block legacy authentication: https://learn.microsoft.com/entra/identity/conditional-access/policy-block-legacy-authentication
• Recommended policies: https://learn.microsoft.com/entra/identity/conditional-access/plan-conditional-access
• Break-glass accounts: https://learn.microsoft.com/entra/identity/role-based-access-control/security-emergency-access
• CA Insights workbook: https://learn.microsoft.com/entra/identity/conditional-access/howto-conditional-access-insights-reporting

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: vinayaklatthe
• Source: vinayaklatthe/microsoft-security-skills
• License: MIT

Install and usage instructions live in the source repository linked above.