The first bounce has a string most operators have never seen: 550 5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records (AS9031519). If you run your own outbound mail — a mailing list, a clinic's appointment reminders, the notification path of a community Nextcloud, a small business invoicing its customers — you may already have it in your logs. The mail it fired on was real. SPF verified. DKIM verified. The authenticated domains matched the visible From domain. Some of it bounced anyway, and the only thing the sending domain was missing is a DNS record that, on any strict reading of Microsoft's published rules, Microsoft had no documented reason to demand from anyone below 5,000 messages a day.
What is verified¶
The record so far, from independent sources with no common ownership:
- November 5, 2026 — MailRoute's outbound logs show the first rejections. The bounce that day carried no
(AS9031519)suffix — a code the next day added. dmarcian dates the onset to the same day. - November 6 — the bounce gains the suffix. Reports begin on the mailop operator list; Neil Jenkins, product chief at Fastmail, describes the same pattern on customer domains with their own send: SPF and DKIM pass, only the DMARC record is absent.
- November 7 — tech press and Microsoft's own Q&A forum fill with reports while Microsoft stays silent.
- November 8 — EasyDMARC reproduces the rejection under controlled conditions: a domain on Zoho Mail with SPF passing and aligned, DKIM passing and aligned, no DMARC record, sending one message to a live.com address. The diagnostic comes back verbatim:
5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records (AS9031519) [MWH0EPF000C6190.namprd02.prod.outlook.com 2026-11-08T14:01:49.694Z …]. The receiving host is worth quoting because it answers the "which Microsoft is this" question:live-com.olc.protection.outlook.com— theolcis Outlook Consumer, the infrastructure behind the free mailboxes. - November 9 — heise's English desk reports the pattern reviewed, including the detail that a Microsoft support-forum moderator has classified the behavior as a disturbance under investigation — Microsoft acknowledges the reports exist, and has still confirmed nothing.
Scope of the recipient side: consumer mailboxes only — outlook.com, hotmail.com, live.com, msn.com and their country variants. Microsoft 365 tenants, as recipients, do not appear to be affected. Scope of the sender side: any domain without a DMARC record, at any volume.
The intermittence is the dangerous part¶
MailRoute's log evidence and EasyDMARC's test agree on one property the announcement-era rules did not have: the rejection is partial. One message from a domain to a Hotmail address delivers; the next bounces; a resend of the bounced one may succeed. That is what makes this hard to notice and worse to debug. The person who sent the message sees a wall of acronyms, assumes a transient glitch, and moves on. The person on the other end never learns anything at all. An appointment confirmation, an invoice, an application — silent non-delivery is indistinguishable from being ignored.
The corollary matters for anyone who will test this: one delivered message proves nothing. A domain is exposed or it is not, based on its DNS, not on the survival of a single probe.
What the rulebook says, and what the bounce does instead¶
Microsoft does have published sender requirements, and they are worth reading against the new bounce, because the distance between them is the story.
| | May 2025 bulk-sender rule | November 2026 rejection (observed) |
| --- | --- | --- |
| Announced | Yes — April 2, 2025 blog post | No — no postmaster notice, no blog, no service-health entry as of November 8 |
| Applies to | Domains sending > 5,000 messages/day | Any domain without a DMARC record, any volume |
| Bounce | 550 5.7.515 — lists the SPF, DKIM and DMARC results | 550 5.7.1 (AS9031519) — names only that records are missing |
| Consistency | Deterministic | Intermittent |
Microsoft's postmaster requirements page still states the 5,000-per-day threshold. Nothing on it describes what small senders have been hitting since November 5.
The honest column. Microsoft has published nothing. Its forum moderator calls it a disturbance being investigated. A smaller set of reports comes from senders whose SPF, DKIM and DMARC are all in place and who still see rejections, sometimes paired with a separate 550 5.7.23 code — those are being treated by the reporting operators as likely misconfigurations or a second, unresolved issue, and this post treats them the same way rather than folding them into the pattern. EasyDMARC's own reading of the main pattern is measured: the behavior is too consistent to be random — domains without DMARC get hit, publishing even p=none fixes it, the bounce text names the missing records — which looks like a rule; whether Microsoft keeps it, softens it, or formalizes it is still open, and the recommendation is to plan as if it stays. Three readings survive the evidence: an unannounced deliberate change, a staged rollout that gets documented later, a filtering misfire. The first two look more likely than the third. The operator homework that follows is identical under all three, which is the practical mercy here.
The fix is one free record, so do it in that order¶
Check the domain first:
dig +short TXT _dmarc.example.org
No output means no DMARC record. A line starting with v=DMARC1 means you have one — in which case this particular rejection is not about you, and the residual 5.7.23 reports belong in a different investigation.
If the record is missing, publish a TXT record at _dmarc.example.org with the value:
v=DMARC1; p=none
Three operational notes, in the order operators get them wrong:
One — the record alone is not the whole fix; alignment is the condition. DMARC passes only when SPF or DKIM passes for the domain in the visible From header: SPF counts via the envelope-from domain, DKIM via the signature's d=. This is Microsoft's own alignment documentation, and it is where most setups that "have authentication" still fail. A DKIM signature from a service's own domain does not align with your From header, no matter how valid it is. Microsoft 365 is worth naming here, twice: it signs with your onmicrosoft.com domain until you enable DKIM for your custom domain, and it does not publish a DMARC record for you. Microsoft's consumer service can bounce mail from Microsoft's own hosted product on an unconfigured domain.
Two — p=none changes how no other receiver treats your mail. It declares that the domain uses DMARC and asks for nothing: no quarantine policy, no reject policy. This is the polite irony — the record that clears the bounce at Outlook.com asserts, in policy terms, exactly nothing. It is a statement of intent: someone owns this domain's authentication. EasyDMARC's criticism of the rushed fix is correct and worth internalizing rather than waving off: a bare p=none with no rua reporting address satisfies Microsoft's check and teaches you nothing about your own mail.
Three — 5xx is forever. A 5.7.1 rejection is permanent from Outlook.com's side: it will not retry the message. Anything that bounced between November 5 and the moment your record went live has to be resent by hand.
For the reporting layer, a self-hosted operator has options the hosted guidance skips: you can point rua=mailto: at an address you control and aggregate the reports yourself — the format is a standardized XML attachment — and you own your delivery logs, which is the evidence layer that turns the reports into something answerable. That entire posture is the subject of the DKIM replay post from last week: DMARC's value to a small operator is not the gate at Microsoft, it is the stream of truth about who sends as your domain, which is exactly the inventory you want before you ever move past p=none.
Why this belongs in a sovereignty newsletter¶
Because of the shape of the curve, not the vendor. In February 2024, Google and Yahoo announced their sender requirements about four months before enforcement, and drew the DMARC line at 5,000 messages a day: below it, SPF or DKIM sufficed, and a small sender with no DMARC was fully compliant. In April 2025 Microsoft announced its version about a month ahead, drew the same line, and enforced it against bulk senders only. In November 2026 consumer Outlook.com began requiring, in effect, a published DMARC record from every sender — without announcement, without threshold, discovered by the people whose mail bounced.
The requirements regime ratchets forward. It is not a scandal that it ratchets — sender authentication is how phishing and business-email-compromise get harder, and a domain that authenticates is harder to impersonate, which is a property worth wanting on purpose rather than under protest. What is worth naming is the information asymmetry at each step: the large providers get the announcement, the integration deadlines, the compliance machinery. Small senders get the bounce. A community list server whose members are at personal addresses finds out from silence on the other end.
The floor has moved and the floor is free. Publish p=none with a rua pointed somewhere you control, read your own reports for a month, and the next step of this curve — whenever it comes for the sub-threshold senders — arrives as a known quantity to you instead of a mystery in your logs. That, and nothing more grandiose, is what this week teaches: the boring DNS inventory is again the whole defense, and on infrastructure you run, it is a file you control.
Sources: MailRoute, "Outlook.com Is Quietly Rejecting Email From Domains Without DMARC" (November 6, 2026); EasyDMARC, "Microsoft Is Now Rejecting Mail From Domains Without DMARC" (November 9, 2026, with the November 8 Zoho→live.com reproduction); dmarcian, "Outlook.com is rejecting some mail from domains with no DMARC record"; heise security / iX (November 9, 2026); mailop list, Fastmail's Neil Jenkins; Microsoft Q&A thread; Microsoft's April 2025 bulk-sender announcement, 5.7.515 documentation and postmaster requirements; DMARC alignment per RFC 7489.