An Entra ID Identity Protection assessment answers three questions: how many accounts Microsoft currently believes are compromised, whether anyone is acting on those signals, and whether a policy forces a response without waiting for a human. Most tenants can answer the first one. Since Microsoft retired the legacy Identity Protection risk policies on October 1, 2026, many can no longer honestly answer the third.
What does Entra ID Identity Protection measure?
Identity Protection scores two different things, and they need two different responses.
| User risk | Sign-in risk | |
|---|---|---|
| Question it answers | Is this account compromised? | Is this sign-in the real user? |
| Typical signals | Leaked credentials, threat intelligence, a pattern of risky sign-ins | Anonymous IP, impossible travel, unfamiliar sign-in properties, token anomalies |
| Where to look | ID Protection > Risky users | ID Protection > Risky sign-ins |
| Recommended response | Require risk remediation (secure password change or session revocation) | Require MFA, every time |
| Clears when | The password is changed or reset, or an admin dismisses the risk | The sign-in completes MFA |
Microsoft's guidance is to keep the two in separate Conditional Access policies. A combined policy fires on whichever condition matches and hides which one did.
What changed on October 1, 2026?
The user risk policy and sign-in risk policy that lived inside Identity Protection were retired. Risk-based access now belongs in Conditional Access only. A tenant that set up its risk policies years ago and never rebuilt them may have no risk enforcement at all today, and nothing in the portal raises an alert about it.
To check, open the Entra admin center and look in two places:
- ID Protection: open the legacy User risk policy and Sign-in risk policy and note whether they were enabled and for whom.
- Conditional Access > Policies: confirm there is an enabled policy with a User risk condition and another with a Sign-in risk condition, covering the same users.
If the first list has scope that the second does not, that gap is unprotected.
How many risky users does your tenant have right now?
In the portal: ID Protection > Risky users, filter Risk state = At risk. With Microsoft Graph PowerShell (the risk APIs need Entra ID P2):
Connect-MgGraph -Scopes "IdentityRiskyUser.Read.All","Policy.Read.All"
# Accounts Microsoft currently considers at risk
Get-MgRiskyUser -All -Filter "riskState eq 'atRisk'" |
Select-Object UserPrincipalName, RiskLevel, RiskLastUpdatedDateTime
# Conditional Access policies that act on risk
Get-MgIdentityConditionalAccessPolicy -All |
Where-Object { $_.Conditions.UserRiskLevels -or $_.Conditions.SignInRiskLevels } |
Select-Object DisplayName, State,
@{n='UserRisk';e={$_.Conditions.UserRiskLevels -join ','}},
@{n='SignInRisk';e={$_.Conditions.SignInRiskLevels -join ','}}
A high-risk user that is still "at risk" weeks later is the finding. It means the signal arrived and nothing happened.
The minimum policy set to keep
- User risk High: Require risk remediation. This grant automatically adds an authentication strength requirement and sign-in frequency "Every time". Password users change their password after MFA; passwordless users have their sessions revoked.
- Sign-in risk Medium and High: Require MFA through an authentication strength, with sign-in frequency "Every time" so an old session cannot satisfy it.
- Exclude break-glass accounts from both, and monitor them separately. See Break-Glass Accounts.
- Make sure users are MFA registered first. A risky user with no registered method cannot self-remediate and is blocked until an admin steps in. Hybrid users also need password writeback.
- Start in report-only mode, read the results for a week, then enable. Microsoft ships templates for both policies.
- Do not forget workload identities. User-scoped Conditional Access does not cover service principals. Risky workload identities need Workload Identities Premium and their own policy. Background in Service Principal Sprawl.
For the wider policy picture, Conditional Access Misconfigurations covers the gaps that sit next to these.
What can each license actually see?
| Capability | Free | P1 | P2 |
|---|---|---|---|
| Risky users report | Medium and high only, no details | Medium and high only, no details | Full, with history |
| Risky sign-ins report | No risk level or detail | No risk level or detail | Full |
| Risk detections report | No | Limited | Full |
| Risk conditions in Conditional Access | No | No | Yes |
| Graph risk APIs, alerts, weekly digest | No | No | Yes |
Some detections also depend on Defender licensing, for example impossible travel and new country come from Defender for Cloud Apps, and suspicious inbox rules from Defender for Office 365 (see Malicious Inbox Rules). A P1 tenant sees that something is risky, but cannot automate the response.
How SecValley checks it
The SecValley Entra ID scan reads risky users, risky service principals and risk detections, and evaluates them against six Identity Protection controls: no high-risk users, medium-risk users kept under a threshold, no risky service principals, no high-risk detections left unresolved, and a Conditional Access policy present for both sign-in risk and user risk. The policy checks read Conditional Access, not the retired legacy policies, so a tenant that lost enforcement on October 1 shows it. Every result is a dated finding with the affected accounts, so a user who stays at risk is visible from one scan to the next.
Frequently Asked Questions
Do I need Entra ID P2 for Identity Protection?
For enforcement, yes. Free and P1 tenants get a limited risky users and risky sign-ins view, but risk conditions in Conditional Access, full risk detail and the Graph risk APIs require P2 (included in Microsoft 365 E5).
What happens to a risky user after a password reset?
The risk state moves from "At risk" to "Remediated" in three cases: the user completes a secure password change after MFA, the user resets the password through SSPR, or an admin generates a temporary password. Dismissing the risk is different: it changes the label but not the password, so a stolen credential still works. Dismiss only after an investigation shows the activity was legitimate.
Should I block high-risk users instead of requiring remediation?
Blocking stops the attacker but also locks out the real user and floods the help desk. Microsoft's recommendation is risk remediation, which lets the legitimate user recover with MFA. Block only where users cannot remediate, such as accounts without MFA registration.
