Microsoft finds spoofed business emails getting through security overrides
Microsoft detected 129 million malicious messages spoofing tracked business-service domains in Q3; tenant and user overrides weakened protection after authentication failed.

Microsoft Threat Intelligence and Defender Security Research detected roughly 129 million malicious messages spoofing the sender domains of tracked business services in Q3 2026. Authentication failed on many of those messages, yet tenant or user security exceptions still allowed delivery. For Microsoft 365 mail administrators, the failure to examine is an allow rule that takes precedence over a malicious verdict.
Microsoft put a sharp number on the problem: “In September, overrides applied to 63% of messages with a recorded security outcome,” mostly tenant-level overrides. The denominator is spoofing messages in that analysis with a recorded outcome, rather than all mail in Microsoft 365. DocuSign was the most frequently spoofed brand in September, accounting for 29% of the tracked business-service spoofing that month. A familiar service name in the visible sender field is a poor reason to trust a message that fails authentication.
How the exception wins
An attacker can claim the sender domain of a real document-signing or survey service. The mail then fails authentication and Microsoft classifies it as malicious, but an organization or user exception changes what happens at delivery. The same exception that keeps a legitimate workflow moving can admit a forged version of that workflow.
Microsoft’s Tenant Allow/Block List documentation describes allow entries as manual overrides of filtering verdicts. Domain and address entries apply to the visible 5322.From address, rather than the envelope 5321.MailFrom. A sender name or apparent From domain therefore needs to be checked against authentication and the policy that handled the message. Microsoft says spoofed-sender allow entries never expire, which makes an old exception worth revisiting after the business reason for it has changed.
The anti-phishing policy documentation places manually overridden spoofed senders on the Spoofed senders tab of the Tenant Allow/Block List. They disappear from the spoof intelligence insight after an administrator changes the verdict. Looking only at that insight can miss the entries an administrator already allowed.
Find delivered forged senders
Start in Defender for Office 365 Threat Explorer with delivered mail claiming the tracked services your staff actually use, especially document-signing, survey and billing services. Compare authentication results, the visible sender, the delivery location and the recorded override for each message. If a malicious verdict was bypassed, preserve the message and investigate clicks and sign-ins before deciding whether anyone acted on it. Use your tenant’s message and identity evidence to decide whether a particular account needs incident response.
Open the Tenant Allow/Block List in the Defender portal and review the Spoofed senders tab as well as domain and address allow entries. Check why each entry exists and who owns the exception. Remove an obsolete or overly broad allow entry that admits impersonation, then confirm on a subsequent legitimate business message that the intended workflow still works. Use the authentication result and delivery or quarantine disposition on subsequent suspect messages to verify that the exception no longer wins. Measure whether the changed policy reduces overridden deliveries in your tenant.
Keep genuine exceptions narrow
Keep legitimate simulation and security-operations mail separate from general sender allow rules. Microsoft’s advanced delivery policy provides explicit phishing-simulation and SecOps-mailbox scenarios with identifiable System override source values in Threat Explorer. Review those configured senders, IPs and recipients against the actual exercise or mailbox rather than granting a blanket service-domain exception. Microsoft says these scenarios bypass several protections, including phishing and spam zero-hour auto purge, so scope them carefully.
The most useful test is a policy and delivery comparison: an approved service message reaches its intended mailbox, while a newly observed sender-identity forgery that fails authentication is handled by the configured anti-phishing policy instead of an obsolete allow entry. Record the exception owner and the changed message disposition.