Microsoft has told affected tenants that a flaw in its Entra Connect Health API let unauthorized callers read tenant metadata and administrator email addresses. The flaw was introduced on 23 July 2026 and stayed live for 119 days before Microsoft deployed a service-side fix on 19 November. Microsoft detected the problem on 9 November, ten days before the fix, and finished notifying impacted organizations by 20 November. BleepingComputer reported the incident on 21 November. The damage was real but bounded: the API returned metadata and contact details, not credentials. No passwords, password hashes, authentication tokens or valid cryptographic keys were exposed, and investigators found no unauthorized configuration changes.
What the flaw exposed
The Connect Health API is the service Microsoft Entra uses to monitor on-premises identity infrastructure, and the flaw sat in the API that returns tenant information. Unauthorized callers could read the responses and pull out tenant IDs, display names, initial domains, country codes, region scopes, license tiers, directory synchronization status, including whether synchronization was enabled and the last synchronization timestamp, and Connect Health onboarding status, along with the email addresses of each tenant’s technical contacts and administrator accounts.
The API responses also returned an expired certificate under the pksCertificate field. The certificate was previously used to register Connect Health agents with a Microsoft internal service that was decommissioned in late 2024. Microsoft said the certificate poses no risk and is no longer used by that service, explaining that it had not been removed from Connect Health storage and continued to be included in the GetTenant API response.
Just as important is what the flaw did not expose. Microsoft stated that the vulnerability did not expose user passwords, password hashes, authentication tokens, valid cryptographic keys, on-premises server and Active Directory topology details, or Azure subscription information. That list bounds the blast radius: the API could only hand back the metadata it was built to return, and nothing credential-grade sat in its responses.
119 days: the timeline
The exposure window ran from a Thursday in July to a Thursday in November. Microsoft’s own account, quoted by SecurityOnline.info in the first detailed report, puts the dates plainly: the vulnerability was introduced on 23 July 2026 and detected on 9 November 2026. The fix followed ten days later, and notifications were complete within a day of the fix. The notification reached the public when administrators shared it on Reddit, and security researcher Dave Kennedy highlighted it on X.
| Date | What happened | Days |
|---|---|---|
| 23 July 2026 | Flaw introduced into the Connect Health API | Day 0 |
| 9 November 2026 | Microsoft detects the vulnerability | 109 days undetected |
| 19 November 2026 | Service-side fix deployed | 10 days after detection |
| 20 November 2026 | Telemetry review finished; impacted organizations notified | 1 day after the fix |
| 21 November 2026 | SecurityOnline.info publishes the first detailed report; BleepingComputer follows | 2 days after the fix |
The 109-day gap between introduction and detection is the number that matters most. The flaw was not a zero-day that burned fast and got patched the same week. It sat in a production API for more than three months while unauthorized callers could query it. Microsoft has not said how many tenants were affected, and has not named the callers, so the scale of the read activity during those 109 days is still unknown.
Why the damage stopped where it did
The long exposure window looks alarming until you compare it with what the API could actually return. A flaw that leaks metadata for 119 days is a different incident from a flaw that leaks credentials for a week, and this one was the former. The exposed fields describe a tenant: its name, its domain, its country, its license tier, its last sync. The exposed email addresses belong to the people who run that tenant’s identity infrastructure. None of that is a password, a token or a key, so none of it can directly sign an attacker in as anyone. But the details could still make a phishing attempt look convincing.
Microsoft verified through internal telemetry logs that the unauthorized activity was read-only. The bug theoretically allowed configuration edits, but forensic logs showed only reads, and investigators confirmed that no unauthorized configuration changes were made. Only tenants that received a direct advisory notice in Azure Service Health had confirmed unauthorized read activity, which means the confirmed impact is narrower than the full population of Connect Health users.
The contrast with Microsoft’s other Entra incident this month sharpens the point. In a separate November campaign, threat actors used the agentic AI tool ARTEX to find exposed application secrets on GitHub, then signed into Entra ID and Azure DevOps with the stolen credentials, reaching Key Vault, SharePoint and file data across 27 organizations, according to Todyl. No Microsoft vulnerability was involved in that campaign. That was real credential abuse. The Connect Health flaw, by comparison, exposed no credentials at all. Both incidents touched Microsoft Entra in November 2026, but they were separate incidents with different mechanisms, and they sit at opposite ends of the damage scale. The difference is what the attacker could actually take.
What should administrators do?
If your organisation uses Entra Connect Health, sign in to the Azure Portal and check Azure Service Health for a direct advisory notice. Only tenants that received one had confirmed unauthorized read activity. Warn your technical contacts and Global Administrators about phishing emails that reference Microsoft Entra, and turn on phishing-resistant multi-factor authentication.
- Check Azure Service Health in the Azure Portal. Only tenants with a direct advisory had confirmed unauthorized reads.
- Warn technical contacts and Global Administrators. Their email addresses were exposed and could attract targeted phishing.
- Turn on phishing-resistant multi-factor authentication. It protects privileged accounts even if an attacker knows the address.
- Run Microsoft’s read-only GetTenant PowerShell script. It shows the types of tenant metadata that could have been exposed.
- Review Entra audit, sign-in and Azure Activity logs. Look for unexpected Connect Health configuration changes or unusual admin activity.
No administrator patching is needed, because Microsoft resolved the flaw on the server side. The work now is defensive. The exposed admin email addresses are the most useful thing the callers obtained, and the realistic follow-on risk is targeted phishing that draws on the exposed names, tenant details and roles to look legitimate. Phishing-resistant multi-factor authentication and Privileged Identity Management are the right controls for privileged accounts here, and they matter more than any password reset, because no password was exposed.
What is still unknown
How many tenants were affected: Microsoft has not disclosed a number, and none should be stated. Who the unauthorized callers were: no threat actor, campaign or attribution has been named. Whether any exposed administrator email address has been used for follow-on phishing: no evidence has been reported, and the phishing risk remains a defensive recommendation rather than an observed event. Whether Microsoft will assign a CVE to the flaw: none has been seen in any source checked. And whether the unauthorized read activity reached beyond the tenants that received direct advisories: Microsoft’s confirmation covers only those tenants.
What to watch next
Whether Microsoft discloses a tenant count or a CVE for the flaw. Whether any affected tenant reports follow-on phishing that uses the exposed admin email addresses. Whether the unauthorized callers are ever identified, and whether the read activity is traced to a specific campaign. And whether the incident changes how Microsoft tests its own identity APIs, given that a flaw in a tenant-metadata endpoint sat undetected for 109 days before Microsoft spotted it. This article will be updated as Microsoft discloses more about the incident.


