Microsoft Patches Maximum-Severity Azure AI Foundry Authentication Gap

En bref

• CVE-2026-85889 affects Azure AI Foundry and carries a CVSS rating of 10.0.
• The root cause is an absent authentication check on a security-critical function.
• An unauthenticated network actor could reportedly gain elevated privileges.
• Microsoft has shipped the fix and states customers need not act.
• Review logs and network exposure around AI Foundry resources as a precaution.

Microsoft has shipped a remediation for a defect in its Azure AI Foundry platform, catalogued as CVE-2026-85889. The issue is an absent authentication control on a function the vendor classifies as security-critical, and the advisory assigns it the maximum CVSS rating of 10.0.

A gap of this kind matters because it removes the credential check that should gate sensitive functionality. According to the source, a remote party with no valid account could reach that functionality and raise its privilege level across the network. Flaws in this class are prized by attackers precisely because they require no credentials and no user interaction to attempt.

Any organization running workloads on Azure AI Foundry inherits the platform's security posture, and that includes K-12 districts and state education agencies standing up AI pilots. Because Azure is delivered as a managed service, the correction is applied on Microsoft's side of the shared responsibility line, and the vendor indicates that tenants have nothing to do.

A 10.0 score reflects maximum technical severity, though practical risk still depends on whether the affected endpoints are reachable from the internet and how the service is configured. Cloud-service vulnerabilities are frequently resolved server-side, which shortens the exploitation window but does not remove the value of checking audit trails for anomalous calls.

What to watch: confirm your AI Foundry resources reflect the patched state, review sign-in and activity logs for unexpected or anonymous access attempts, and track any revision to the vendor's guidance. Publication of proof-of-concept details after a patch often precedes a rise in scanning and exploitation attempts, so continued monitoring is warranted.

À faire maintenant

1. Confirm that Azure AI Foundry resources in your tenant report the patched service state; open a support case if the status is unclear.
2. Hunt through Azure Activity, Entra sign-in, and diagnostic logs for anonymous or unexpected calls to AI Foundry endpoints before the fix landed.
3. Reduce network reachability by applying private endpoints, IP allowlists, and firewall rules so AI workspaces are not broadly exposed.
4. Enforce least privilege on service principals and managed identities tied to AI workloads, and rotate any keys or tokens that may have been exposed.
5. Enable diagnostic logging for AI Foundry and route it to your SIEM, alerting on role-assignment and privilege-escalation events.
6. Monitor the Microsoft advisory and CVE-2026-85889 for updates, and reassess if the vendor changes its no-action guidance.
7. Brief application owners and AI pilot teams on what to report if they observe anomalous behavior in these services.

Références CVE

• CVE-2026-85889

Source originale

The Hacker News