On this page
- What is Direct Send, and why does anyone use it?
- What does RejectDirectSend do, and where does the blank field come in?
- Who needs to act, and who does not?
- How do I restrict Direct Send this week?
- How do I tell if a message used this?
- Am I affected? Where you stand right now
- What if a message already got through?
- What can a website scan tell me here?
- What we could not confirm
Microsoft 365 has a setting called RejectDirectSend that blocks anonymous mail sent as your own domain straight to your tenant. ReliaQuest reported on 3 September 2026 that a message with an empty envelope sender gets past it. Keep the setting on, restrict Direct Send to known IP addresses with an inbound connector, and hunt your own mail for empty-sender messages that show an internal From address.
| Fact | Detail | Source |
|---|---|---|
| What was reported | An empty SMTP envelope sender bypasses RejectDirectSend; ReliaQuest says no credentials or lookalike domain are needed | ReliaQuest, 3 Sep 2026 |
| The setting | RejectDirectSend on Set-OrganizationConfig; $false means Direct Send is not blocked; Exchange Online only | Microsoft Learn, checked 24 Oct 2026 |
| Microsoft's position | ReliaQuest says Microsoft was made aware through multiple attempts. The report describes no fix. Microsoft's Direct Send page says it is working on an option to disable Direct Send by default | ReliaQuest; Microsoft Learn, checked 24 Oct 2026 |
| CVE or KEV entry | None that we found at our check on 4 November 2026 | this page |
| Who is affected | Microsoft 365 tenants that accept mail at their own MX endpoint, which is every tenant by default | Microsoft Learn |
| Exploitation | ReliaQuest says it saw many instances of the empty-envelope pattern over the past year, aimed at executives and business roles. Vendor-claimed | ReliaQuest |
This page covers one narrow question: what the bypass is, and what an admin can do about it this week. The wider list of October incidents is in the October 2026 tracker, and more mail-security pages sit with the other email guides.
What is Direct Send, and why does anyone use it?
Direct Send is the way a printer, scanner or old application sends mail to your own staff without logging in. The device connects to your tenant's MX endpoint on port 25 and sends as an address in one of your domains. Microsoft's page lists the requirements: no authentication, port 25, and an address from an accepted domain.
Microsoft's own wording is cautious. It says most customers do not need Direct Send, and recommends it only for advanced customers who accept the duties of an email server admin. It also says messages sent this way are treated like anonymous email from the internet, so all scanning and protections apply.
That last sentence is the point. Your tenant must accept unauthenticated mail at the MX endpoint, because that is how every outside sender reaches you. Direct Send is the same door, used by your own gear.
| Method | Authentication | Who it suits | Checked |
|---|---|---|---|
| SMTP AUTH (client submission) | A mailbox username and password, or OAuth | Apps that can log in; Basic auth is scheduled for deprecation | 4 Nov 2026 |
| SMTP relay | Inbound connector, certificate or fixed IP | Devices with a fixed IP or a certificate | 4 Nov 2026 |
| Direct Send | None | Legacy devices that support nothing else; internal recipients only | 4 Nov 2026 |
| High Volume Email | HVE account, password or OAuth | Large volumes to internal recipients | 4 Nov 2026 |
Source for the table: Microsoft Learn device and application guide, updated 5 October 2026.
What does RejectDirectSend do, and where does the blank field come in?
Microsoft's cmdlet reference describes RejectDirectSend as the switch that blocks Direct Send. When it is $true, Exchange Online rejects anonymous messages sent from your own domain to your mailboxes, if they match no inbound connector (only connectors that match the sender IP or a certificate count) and the domain in the MAIL FROM address is one of your accepted domains. When it is $false, Direct Send is not blocked.
ReliaQuest Threat Research (Tristano Di Liberto and Austin Ritchie) reports that the control evaluates the domain in the SMTP envelope sender. If that field is empty, there is no domain to check, and the message passes the control. The visible From address, which is what a person reads in Outlook, is a different field and can name anyone, including your CFO.
An empty envelope sender is not unusual by itself. RFC 5321 uses a null reverse-path on notification messages so that bounce messages do not trigger more bounces. That is why a blanket block on empty senders is the wrong fix: it would also catch genuine delivery failure notices. Microsoft's header documentation even has reason codes 501 and 502 for valid non-delivery reports that skip DMARC enforcement.
Who needs to act, and who does not?
Any organisation with Microsoft 365 mail flow through Exchange Online can be on the receiving end, because accepting mail at the MX endpoint is how the service works. The people who need to act are the ones who administer the tenant: an IT team, a managed service provider, or the one person who set it up.
If your mail is hosted elsewhere, this report does not describe you. The report covers Microsoft 365 and the RejectDirectSend setting in Exchange Online, and we found no primary source about other providers.
Agencies should check each client tenant they manage. The question for each is the same: who sends mail as this domain without logging in, and from which IP address.
How do I restrict Direct Send this week?
Work in this order. Each step is reversible, and the first one is a list, not a change.
Step 1: list what really sends unauthenticated mail
- Printers, scanners and copiers that email scans
- Line-of-business applications and monitoring tools that email alerts
- Website contact forms and plugins that send as your domain through your own tenant (see why form emails go missing)
- The public IP address each of them sends from
Message trace helps. In the Exchange admin center message trace, set a range up to 90 days, and look at the original client IP and the connector ID for mail from your own addresses to your own mailboxes.
Step 2: keep or turn on RejectDirectSend
Microsoft documents the parameter as a boolean on Set-OrganizationConfig.
Set-OrganizationConfig -RejectDirectSend $true
Turn it on only after step 1 and step 3 are ready for any device that still needs to send. Mail from a device that has no connector will start failing once the setting is on.
Step 3: give each legitimate sender a connector
ReliaQuest's first recommendation is an IP-restricted inbound connector that permits only the inventoried devices and applications that need unauthenticated sending, with RejectDirectSend kept enabled. Microsoft's guide describes the admin center route: Mail flow, Connectors, Add a connector, connection from "Your organization's email server", and authentication by the sending IP address, which must belong exclusively to your organisation.
The New-InboundConnector reference says SenderIPAddresses applies to Partner type connectors when RestrictDomainsToIPAddresses is $true, which then rejects mail from the listed sender domains arriving from any other IP. A certificate is the stronger option where the device supports it. Microsoft's connector guide explains how connectors are matched.
Step 4: move what can move
Where a device or application can log in, move it to SMTP AUTH with OAuth, or to High Volume Email for internal recipients. Microsoft's guide notes that Basic authentication for client submission is scheduled for deprecation, so a new setup should not depend on it.
Step 5: review every override
ReliaQuest asks defenders to review every setting that can override a filtering verdict. Microsoft lists them: allowed senders and domains in anti-spam policies, the Tenant Allow/Block List, mail flow rules that set the spam confidence level, and users' safe sender lists. The Tenant Allow/Block List page says allow entries skip the filters tied to the entity they name, and that they apply to the From address. An allow entry for your own domain is exactly the kind of entry that helps a forged internal From address.
Also confirm that spoof intelligence and the Honor DMARC record policy setting are on, and that your own domain publishes a DMARC policy. RFC 7489 says DMARC authenticates use of the From domain, which is the field this bypass leaves free to forge. A passing check is not always a safe message either, as the iCloud spoofing case shows, and when your own legitimate mail fails, DKIM and DMARC alignment is the usual cause.
How do I tell if a message used this?
ReliaQuest lists these signals for a message of this kind. Each is a clue, and none is proof alone.
| Signal | Where to see it | What it suggests |
|---|---|---|
Empty Return-Path (<>) | Message headers; return_path in the message trace Extended report | The envelope sender was empty |
| From address in one of your accepted domains | Message headers | The message claims to be internal |
| SPF, DKIM or DMARC failed | Authentication-Results header | The claim is not backed up |
X-MS-Exchange-Organization-AuthAs: Anonymous | Message headers | No authentication on the connection |
| High spam confidence (SCL 9) | X-Forefront-Antispam-Report | Filtering scored it as spam |
Microsoft notes that in cloud tenants the SCL value does not decide the action taken. Read the category and the delivery location instead.
For a search across your mail, Defender for Office 365 customers can use the EmailEvents table, which has separate columns for the envelope sender (SenderMailFromAddress), the From address (SenderFromAddress), the From domain and the authentication results. Microsoft's schema page says the table needs Defender for Office 365.
EmailEvents
| where isempty(SenderMailFromAddress)
| where SenderFromDomain in~ ("yourdomain.com")
| where AuthenticationDetails has "fail"
| project Timestamp, SenderFromAddress, RecipientEmailAddress, Subject, DeliveryLocation, AuthenticationDetails
Replace the domain with your own accepted domains. A valid bounce can appear in the results too, so read the subjects and the delivery location. A hit means a message showed this pattern. It does not tell you whether anyone acted on it.
Am I affected? Where you stand right now
Am I affected? Where you stand right now
A scan from outside cannot read your tenant settings, your connectors or your message history. These states come from what you can show from inside the tenant.
- Detected A message in your logs has an empty envelope sender, a From address in your own domain and failed authentication. Treat it as an attempt that reached your tenant. Find the recipient, check whether it was delivered, and follow the incident steps below.
- Unknown You have not read the RejectDirectSend value, listed your connectors, or searched your mail for the pattern. This is the common case. Do the three reads in the checklist below.
- Confirmed by owner evidence You hold the setting value, a connector list limited to named IP addresses, a reviewed list of overrides, and a hunt result over your full retention window. You have shown what your tenant looks like today. Microsoft has not described a fix for the bypass, so repeat the hunt on a schedule.
No scan, ours included, moves you from "unknown" to "confirmed".
Evidence checklist to hold on file:
- Output of
Get-OrganizationConfig | Format-List RejectDirectSend, with the date - The list of inbound connectors and the IP addresses each one allows
- The list of allowed senders, allowed domains, Tenant Allow/Block List entries and mail flow rules that set spam confidence
- The hunt query, its date range and its result count
What if a message already got through?
A forged internal message is a social engineering problem first. Someone may have read it, clicked, replied or paid.
- Find the recipients and what each one did. Ask directly, without blame.
- If a link was opened or a file was run, follow the general incident containment steps, and reset the passwords involved.
- If an invoice or payment instruction was followed, contact your bank the same day.
- Pull the original message headers and keep them with your notes, so you can check them against the signals above.
- Decide who in the company is told, and if customers were affected, a notice template saves time. The responsibility page helps when several parties run parts of your setup.
Message trace data is limited. Microsoft's page says you can query up to 90 days, in steps of 10 days at a time, and that results beyond 10 days come as a downloadable report that can take hours. Start the request early.
What can a website scan tell me here?
Not much about this issue. Red's free scan reads the stack your site serves and matches each version to known vulnerabilities. It is a lookup. It does not read your Microsoft 365 tenant, your connectors or your mail logs, and this bypass is not a version of anything it can match. The limits of an outside view are set out in this note on what outside scans miss.
If your question is whether the fix you applied took effect, the same rule holds as for any fix: only evidence from the system itself counts. A guide to proving a fix actually worked covers what that evidence looks like.
What we could not confirm
At our check on 4 November 2026 we could not confirm these from a primary source, so the page does not state them:
- Whether Microsoft has changed RejectDirectSend to close the empty-sender path, or has a date for doing so. ReliaQuest describes no fix, and Microsoft's Direct Send page says only that an option to disable Direct Send by default is in progress.
- How many tenants were affected. The count is not public.
- Whether every connector configuration blocks the path. ReliaQuest tested an IP-restricted connector, so test your own in your tenant and record the result.
Questions people ask
What is the RejectDirectSend bypass?
ReliaQuest Threat Research reported on 3 September 2026 that a message with an empty SMTP envelope sender gets past the RejectDirectSend control in Microsoft 365. The control checks the domain in the envelope sender, and an empty value leaves nothing to check. The visible From address can still show a person inside your company.
I turned on RejectDirectSend. Am I covered?
The setting is worth keeping on, but the report says it can be bypassed, and it did not describe a Microsoft fix. Treat it as one layer. Add an IP-restricted inbound connector for the devices that really need to send, review every filtering override, and hunt your own mail for the pattern.
How do I tell if someone has already used this on my tenant?
Look for messages where the envelope sender is empty but the visible From address is one of your own domains, and where SPF, DKIM or DMARC failed. Message trace and the EmailEvents table both show these fields. Run the search over the longest range your retention allows.
Will turning off Direct Send break my scanner or my website forms?
It can. Printers, scanners and older applications that send as an unauthenticated mail server are the usual users. List them first, give each a fixed outbound IP, and allow only those IPs through a connector before you tighten anything.
Does this affect Google Workspace or other mail providers?
The report describes Microsoft 365 and the RejectDirectSend setting in Exchange Online. We did not find a primary source about the same behavior at other providers, so this page makes no claim about them.
Can a website scan tell me if my tenant is exposed?
No. This setting lives in your Microsoft 365 tenant and in mail logs, which an outside scan cannot see. Only the admin who runs the tenant can read the setting, the connectors and the message history.
Sources
- ReliaQuest Threat Research: one blank field bypasses the Direct Send control (3 September 2026)The finding: an empty SMTP envelope sender bypasses RejectDirectSend. Detection signals and mitigations. No update notes at our 4 November 2026 check.
- Set-OrganizationConfig: RejectDirectSend parameter (Microsoft Learn)What the setting blocks and the connector and envelope-domain conditions. Exchange Online only. Checked 24 October 2026.
- Set up a device or application to send email with Microsoft 365 (Microsoft Learn)Direct Send, SMTP relay, SMTP AUTH and High Volume Email compared. Page updated 5 October 2026; checked 4 November 2026.
- New-InboundConnector (Microsoft Learn)Connector types and the sender IP restriction parameters. Checked 24 October 2026.
- Set up connectors to route mail (Microsoft Learn)How inbound connectors identify your own servers, and connector selection order.
- Message trace in the new Exchange admin center (Microsoft Learn)Time ranges, the Extended report and its return_path field. Page updated 5 October 2026.
- EmailEvents table in advanced hunting (Microsoft Learn)Envelope sender, From address and authentication columns. Needs Defender for Office 365.
- Anti-spam message headers (Microsoft Learn)Authentication-Results fields, compauth reason codes, and what SCL means in cloud tenants.
- Anti-phishing policies in Microsoft 365 (Microsoft Learn)Spoof intelligence, Honor DMARC record policy, and the ways an admin can override them.
- Set up DMARC in Microsoft 365 (Microsoft Learn)How DMARC aligns the envelope and From domains, and how inbound failures are handled.
- Tenant Allow/Block List (Microsoft Learn)Allow entries skip filters for the entity they name. Entries apply to the From address.
- RFC 5321: Simple Mail Transfer ProtocolWhy an empty envelope sender (null reverse-path) is normal for bounce messages.
- RFC 7489: DMARCDMARC authenticates use of the RFC5322.From domain.
Facts on this page are dated and point at the primary source next to them. If a vendor status changes after the date shown, the vendor's page wins; tell us at support@run.cloud and we will update it.