Two Defender flaws under active exploitation, Microsoft confirms

A SYSTEM-level link-following bug and a denial-of-service issue in Microsoft Defender are both being abused in the wild.

ThreatVectr Newsdesk· 2 min read
Two Defender flaws under active exploitation, Microsoft confirms
Share

Microsoft has acknowledged that two vulnerabilities in its own security product are being exploited in the wild, which is the sort of disclosure one reads twice before continuing.

The more serious of the pair, CVE-2026-41091, is a local privilege escalation rated 7.8 on CVSS. The root cause is improper link resolution before file access — the classic link-following class of bug — and a successful exploit hands the attacker SYSTEM. Predictably, that is the ceiling on a Windows endpoint, which makes Defender itself a useful ladder for anyone who already has a foothold.

The second issue is a denial-of-service flaw in the same product. Microsoft has flagged it as exploited but has been characteristically sparing on detail. No indicators. No threat actor named. No advice beyond patching.

Both flaws require local access, so this is not a wormable, internet-facing emergency. It is the quieter sort of problem. An initial-access broker drops a loader, the loader needs SYSTEM to disable telemetry or install a driver, and Defender obligingly provides the lift. Turning the EDR into the escalation primitive is a tidy bit of work.

Microsoft says the Defender platform updates ship through the usual channels and that most environments will pick them up automatically via the Malware Protection Platform. Admins running locked-down or air-gapped fleets, which is to say a non-trivial slice of regulated industries, will need to verify the platform version manually. The fixed build numbers are listed in the MSRC advisory.

There is no public proof-of-concept yet. That tends not to last. Link-following bugs in privileged Windows services have a well-trodden reverse-engineering pipeline, and once the patch diff is out, working exploits usually follow within days. James Forshaw of Google Project Zero has been documenting this exact bug class for the best part of a decade, and the recipe is not a secret.

What is mildly awkward for Redmond is the framing. Defender is the control that is meant to catch the post-exploitation behaviour these flaws enable. When the control becomes the primitive, the assurance story gets harder to tell (the marketing team will manage).

The practical guidance is unglamorous. Confirm the platform version on every endpoint. Treat any host that has lagged on Defender updates as a candidate for review. And if your detection logic assumes Defender is tamper-resistant from a local administrator context, revisit that assumption.

The wider question — why a security product keeps shipping link-following bugs in 2026 — is one for another week.

© 2026 Threat Vectr