Outlook Zero-Day Under Active Exploitation: What We Know and How to Respond

A previously unknown Outlook vulnerability is being exploited in the wild right now. Here is what defenders need to understand about the attack path, the risk to Exchange and Microsoft 365 estates, and the mitigations that actually reduce exposure.

A zero-day in Microsoft Outlook is being exploited in the wild, and for defenders the uncomfortable part is not that another mail client bug exists — it is that Outlook sits at the exact point where untrusted content, rich rendering, and a user's identity all meet. When a flaw in that layer is weaponised before a patch exists, the attacker gets a head start measured in days or weeks, and every mailbox in the organisation becomes a candidate entry point.

Why Outlook keeps attracting attackers

Outlook is not just an email reader. It is a rendering engine for HTML, a parser for calendar invites, a handler for embedded objects, and a client that talks to Exchange, to Microsoft 365, and to a long list of COM components on the endpoint. That surface area is precisely what makes it valuable to an attacker: a single crafted message can reach a user with no link to click and, depending on the flaw, sometimes no obvious prompt at all.

The pattern is familiar from the last several years of Outlook history. The 2023 Outlook for Windows elevation-of-privilege issues around the MAPI interface, the Moniker handling problems, and the well-documented CVE-2023-23397 — a zero-click NTLM credential leak triggered by a calendar reminder — all share the same underlying characteristic. The vulnerability lived in how Outlook processed a piece of structured content, not in whether the user was careless. That matters for how you defend: awareness training is not a control against this class of bug.

When a new Outlook zero-day is confirmed as exploited, treat the initial disclosure as the start of an investigation rather than the end of one. The public advisory tells you the affected component and the vendor's severity rating. It rarely tells you the exploitation pattern, the delivery mechanism, or whether your particular configuration is in the blast radius.

Anatomy of the attack path

Most Outlook exploitation chains follow a recognisable sequence, and understanding it helps you decide where to place detection and containment. The attacker first delivers a crafted message — an email, a calendar invite, or a shared item — to the target mailbox. Delivery can come from an external sender, from a compromised partner tenant, or from an already-compromised internal account, which is why sender reputation alone is not sufficient.

The message is then processed by the client. Depending on the vulnerability, exploitation may occur on preview, on open, or on a background synchronisation event. The classic zero-click pattern — where merely receiving the item is enough — is the most dangerous because it removes the last human decision point from the chain. From there, the attacker's objective is usually one of three things: leak credentials (NTLM relay or hash capture), achieve code execution in the context of the user, or establish persistence through a malicious add-in or a crafted profile configuration.

Post-exploitation is where the real damage happens. A successful credential leak feeds relay attacks against internal services that still accept NTLM. Code execution in the user context enables reconnaissance of the mailbox, lateral movement via delegated permissions, and the classic business email compromise playbook: inbox rules, forwarding, and OAuth consent abuse. In our penetration tests, we routinely find that once an attacker holds a valid mailbox session, the rest of the estate is far more reachable than the perimeter suggests.

Important: If the advisory describes a zero-click or preview-triggered flaw, assume that opening the message is not required for compromise. In that scenario, mail filtering and client hardening are your primary controls, not user behaviour.

What the advisory tells you — and what it doesn't

Vendor advisories for Outlook zero-days typically give you the CVE identifier, a CVSS base score, the affected product versions, and a workaround or patch. What they frequently omit, at least in the first revision, is the exploitation status per version, the exact trigger condition, and whether the flaw is reachable through Exchange on-premises, Microsoft 365, or both. That gap is where a lot of organisations misjudge their exposure.

Use the following as a triage frame while you wait for fuller technical detail. It is deliberately generic because the specifics of this disclosure are still firming up, but the decision logic holds for any Outlook client vulnerability.

QuestionWhy it mattersAction if yes
Is the client in scope?Version and channel determine patch availabilityInventory builds, prioritise exposed ones
Is it zero-click?Removes the user decision pointFilter at the gateway, harden preview
Does it leak credentials?Feeds relay and lateral movementBlock outbound NTLM, audit relay paths
Is on-prem Exchange involved?Changes patch and mitigation orderPatch server and client together

The CVSS score is a starting point, not a verdict. A 7.8 that requires user interaction is a different operational problem from a 7.8 that fires on preview. For a mail client, the difference between those two is often the difference between an incident and a near miss.

Detection: what to look for now

Detection for Outlook exploitation leans heavily on telemetry you may already have but are not correlating. Start with process ancestry. Outlook spawning rundll32.exe, regsvr32.exe, mshta.exe, powershell.exe, or wscript.exe is almost never legitimate in a normal business environment. The same applies to OUTLOOK.EXE making outbound SMB or WebDAV connections — that is the signature of a credential-leak primitive being triggered.

On the network side, watch for authentication attempts originating from user workstations against internal file shares or Exchange endpoints that do not fit the user's normal pattern. NTLM relay activity often shows up as a burst of failed authentications followed by a success from an unexpected source. If you have 4624 and 4625 events centralised, build a correlation around workstation-to-server NTLM in a short window.

On the mailbox side, the post-exploitation signals are well understood: new inbox rules that forward or delete, new delegate permissions, new OAuth applications granted mail access, and sign-ins from unusual locations or user agents. Microsoft 365 audit logs expose these, but only if you are actually alerting on them rather than archiving them.

# Quick triage: Outlook spawning suspicious children (Sysmon)
EventID: 1
ParentImage: *\OUTLOOK.EXE
Image: *\rundll32.exe OR *\regsvr32.exe OR *\mshta.exe OR *\powershell.exe

If you run Exchange on-premises, add a check for messages that were delivered but never opened, and for calendar items with unusual property lengths or embedded references. These are the artefacts that tend to survive after the payload has fired.

Remediation and hardening

Patch as soon as the vendor ships a fix, but do not treat patching as the whole response. The window between disclosure and patch is exactly when attackers operate, and the controls that reduce blast radius during that window are the same ones that make your estate resilient afterwards. Work through these in order of impact.

  • Inventory and patch. Identify every Outlook build in use, including Click-to-Run channels, LTSC installs, and any client still talking to on-prem Exchange. Patch the internet-facing and executive populations first.
  • Block outbound NTLM. Disable NTLM where possible and enforce SMB signing and LDAP signing. This neutralises credential-leak primitives that rely on coercing authentication to an attacker-controlled host.
  • Harden the client. Disable preview panes for untrusted senders where policy allows, restrict add-ins to an approved list, and enforce Protected View and Attachment Manager settings via Group Policy.
  • Filter at the gateway. Block or quarantine messages with malformed calendar properties, external sender spoofing, and unusual embedded objects. Rules that target the delivery mechanism buy you time even before a signature exists.
  • Reduce identity blast radius. Enforce MFA, conditional access, and least-privilege mailbox delegation. Audit OAuth consent grants and revoke anything the business cannot justify.
  • Hunt, then verify. Run the detection queries above across the last 30 days, not just since disclosure. If the flaw was exploited before it was public, the evidence is already in your logs.
Caution: Disabling preview or blocking add-ins can break legitimate workflows. Pilot the change with a small group and communicate it, or you will spend the incident window arguing with the business instead of containing the threat.

The bigger lesson for your estate

Outlook zero-days are not rare, and they are not going away. The mail client is a permanent high-value target because it combines identity, content rendering, and endpoint code in one process. The organisations that come through these events well are not the ones with the best threat intelligence feed — they are the ones who already knew what was in their estate, had NTLM relay paths closed, and could answer the question "which mailboxes were touched?" within an hour.

From a penetration-testing perspective, the most useful thing you can do this week is not wait for the next advisory. It is to test whether a single compromised mailbox in your environment leads to domain-wide access. In our experience, it usually does — through delegated permissions, unconstrained delegation, or service accounts that trust NTLM. Fixing that chain is worth more than any single patch, because it is the difference between one exploited zero-day and a full breach.

Treat this disclosure as a prompt to review the fundamentals: patch cadence for client software, authentication protocols that should have been retired years ago, and detection coverage for the post-exploitation behaviours that follow every successful phish or exploit. The vulnerability will be patched. The architectural weaknesses it exposes will not fix themselves.

All articles