Skip to content
Back to guides
Guide

Allow a phishing simulation in Microsoft 365 (Advanced Delivery): 2026 guide

Allow a third-party phishing simulation in Microsoft 365 with the Advanced Delivery policy: domains, IPs, IPv6 in PowerShell, High Confidence Phish fixes, how to check. Updated November 18, 2026.

Thomas Ferreira 13 min read

To allow a third-party phishing simulation in Microsoft 365, add the tool to the Advanced Delivery policy in the Microsoft Defender portal, Phishing simulation tab: at least one domain, the envelope sender or DKIM domain, and at least one sending IP address. It is the only mechanism Microsoft provides for these messages, and it works on every cloud mailbox, Business Premium included, with no Defender for Office 365 Plan 2 licence.

Updated November 18, 2026. Every setting below was checked on Microsoft Learn that day, mainly on Configure the advanced delivery policy (updated July 3, 2026) and Report phishing and suspicious emails in Outlook (July 3, 2026). The method is the same for any simulation tool: GoPhish, KnowBe4 or nophi.sh.

nophi.sh values for Advanced Delivery

  • IPv4 sending IP: 85.208.144.52 (enter it in the portal)
  • IPv6 sending IP: 2a09:8740:0:3::52 (PowerShell only)
  • Domain: the sending domain of your campaigns and, if different, the envelope sender domain. Both are shown in the console, Settings › Sending simulations, because they follow the sending domain chosen for your campaigns.
  • Simulation URLs (optional): app.nophi.sh/t/*

The console detects your mail provider from your MX records, shows these values and the PowerShell command ready to copy, then sends you a real test email through the same path as the simulations. You confirm delivery with one click on “I received this email”.

Why Microsoft 365 quarantines phishing simulations

A good simulation email looks exactly like real phishing, and Exchange Online Protection and Defender for Office 365 are built to stop it. Most of the time the message is classified as “High confidence phish” and goes to quarantine, where the employee cannot release it.

For these messages, Microsoft applies its secure by default rule: “Microsoft 365 doesn’t allow allowlists or filtering bypass for messages that are identified as malware or high confidence phishing”. A mail flow rule that sets SCL to -1, the IP Allow List of the connection filter or a URL allow entry in the Tenant Allow/Block List will not rescue a simulation classified that way. Microsoft says so for the last one: don’t use URL allow entries, use the advanced delivery policy (Tenant Allow/Block List, July 31, 2026). The Secure by default page names phishing simulations among the cases this exception is for.

Which licence do you need for Advanced Delivery?

You need no extra Defender licence. The Microsoft Learn page says the policy applies to “Built-in security features for all cloud mailboxes” and to Defender for Office 365 Plan 1 and Plan 2. Microsoft 365 Business Premium, which includes Plan 1, and Microsoft 365 E3 are therefore fine.

What you need are permissions:

  • in Email & collaboration RBAC, the Security Administrator role group, and in Exchange Online, the Organization Management role group;
  • or, with Defender XDR unified RBAC, the “Authorization and settings/Security settings/Core Security settings (manage)” permission;
  • or an Entra role: Security Administrator, or Global Administrator (which Microsoft advises keeping to a minimum).

Not to be confused with Attack Simulation Training, Microsoft’s own simulation tool, which does require E5 or Plan 2. Our page on Attack Simulation Training licences and cost covers that.

Configure Advanced Delivery for a third-party phishing simulation

  1. Sign in to security.microsoft.com with an account that has the permissions above.
  2. Go to Email & collaboration, Policies & rules, Threat policies, then Advanced delivery, in the Rules section. The direct address is https://security.microsoft.com/advanceddelivery.
  3. Select the Phishing simulation tab, then Add, or Edit if an entry already exists.
  4. In Domain, add the envelope sender domain, also called MAIL FROM, 5321.MailFrom or P1 sender, the one shown in the Return-Path header of the received message, or the DKIM domain given by the vendor. The domain shown in the “From” field is not enough if it differs. Up to 50 entries.
  5. In Sending IP, add at least one IPv4 address: a single IP, a range (192.168.0.1-192.168.0.254) or a CIDR block (192.168.0.1/25). Up to 10 entries. Use the narrowest range you can.
  6. Simulation URLs to allow: optional. Microsoft says this setting isn’t required for links in email phishing simulations; it is for links sent in Teams messages or Office documents. Up to 30 entries.
  7. Save.

Microsoft’s matching rule is simple: “There must be a match on at least one Domain and one Sending IP, but no association between values is maintained”. A message is recognised when it matches at least one domain and at least one IP on the list, with no link required between the two.

Microsoft gives no propagation time for this policy. Send a test after saving, and keep in mind that messages received before stay in quarantine.

Add IPv6 in PowerShell (New-ExoPhishSimOverrideRule)

The portal only takes IPv4: “Currently, IPv6 addresses are supported only in PowerShell”. If the simulation tool also sends over IPv6, as nophi.sh does, use Exchange Online PowerShell:

Connect-ExchangeOnline

# First setup: the policy, then the rule
New-PhishSimOverridePolicy -Name PhishSimOverridePolicy
New-ExoPhishSimOverrideRule -Policy PhishSimOverridePolicy -Domains your-sending-domain.com -SenderIpRanges 85.208.144.52,2a09:8740:0:3::52

Replace your-sending-domain.com with the domain shown in your tool’s console. Whatever name you pass to -Name, the policy is always called PhishSimOverridePolicy.

If a rule already exists, for example one created in the portal, just add the addresses:

Get-ExoPhishSimOverrideRule | Set-ExoPhishSimOverrideRule -AddSenderIpRanges 85.208.144.52,2a09:8740:0:3::52

To add a domain to an existing rule, the Set-ExoPhishSimOverrideRule reference page uses -AddDomains (and -RemoveDomains); the general article on the policy shows -AddSenderDomainIs, a name the reference page does not list. Follow the cmdlet page. It also caps domains at 20 in PowerShell, against 50 in the portal.

Simulation URLs, if you decide to add them, go into the Tenant Allow/Block List with the AdvancedDelivery subtype:

New-TenantAllowBlockListItems -Allow -ListType Url -ListSubType AdvancedDelivery -Entries "app.nophi.sh/t/*" -NoExpiration

What the policy changes for simulation messages

For a message recognised as a phishing simulation, Microsoft Learn describes this behaviour:

  • filters in Microsoft 365 take no action on the message;
  • zero-hour auto purge (ZAP) for spam and phishing does not remove it afterwards;
  • Safe Links still wraps the links, but does not block or detonate them;
  • Safe Attachments does not detonate attachments;
  • default system alerts are not triggered, and automated investigation and response (AIR) ignores the message;
  • if an employee reports it with the built-in Report button, no alert, investigation or incident is created, but the message appears on the User reported tab of the Submissions page.

The policy does not replace the authentication of the sending domain: a serious simulation tool sends from an authenticated domain, with SPF, DKIM and DMARC aligned. Our SPF, DKIM and DMARC guide explains what those records check.

Why simulations stay quarantined as “High Confidence Phish”

If messages are still blocked after setup, work through this list in order. Each point comes from the Microsoft Learn page of the policy.

  1. The domain doesn’t match. Open the source of a quarantined message. The domain you added must be the Return-Path (MAIL FROM) domain or the d= of the DKIM signature, not the “From” domain.
  2. The IP doesn’t match. Microsoft compares the connecting IP it sees. An SMTP relay, NAT or intermediate sending service changes that IP: read the first Received header added by Microsoft.
  3. A gateway filters before Microsoft 365. If your MX does not point to Microsoft 365 (Proofpoint, Mimecast, Barracuda in front of the tenant), the IP in the Authentication-Results header must match the one in the policy. Otherwise, turn on Enhanced Filtering for Connectors so Microsoft finds the original IP. Allow the tool in the gateway itself too.
  4. The message makes a round trip. Enhanced Filtering does not work when mail goes from the internet to Microsoft 365, then to an on-premises server or third-party service, and back to Microsoft 365. Microsoft asks you not to work around this by adding the IPs of that intermediate infrastructure.
  5. The simulation is internal (DIR:INT) or goes through Exchange Server in hybrid. This is not supported. Microsoft offers two workarounds: a dedicated Receive connector that doesn’t authenticate the simulation messages as internal, or sending straight to the tenant’s Microsoft 365 MX (yourdomain-com.mail.protection.outlook.com).
  6. The tool drops messages by Direct Injection. “Direct Injection emails bypass transport, so Advanced Delivery policy does not apply to them.”
  7. The message arrived before setup. It stays in quarantine: release it or send again.

Shared mailboxes and the built-in Outlook Report button

Microsoft has put the Report Message and Report Phishing add-ins in maintenance mode, to be retired eventually: the built-in Report button in Outlook takes over. For shared mailboxes, Microsoft Learn is specific: in Outlook for Mac, Outlook for iOS and Android, the new Outlook for Windows and Outlook on the web, the built-in button can report a message from a shared mailbox or a delegated mailbox. The delegate needs Send As permission on the mailbox; without it, “the reported message is not sent to the reporting mailbox”, it is only removed from the folder. Classic Outlook for Microsoft 365 is not on that list.

Three rules so reports count:

  • the reporting mailbox must be an Exchange Online mailbox in your organisation: distribution groups and external addresses are not accepted (User reported settings, September 28, 2026);
  • add that mailbox on the SecOps mailbox tab of the Advanced Delivery policy. Microsoft explains that otherwise “a user reported message might trigger a training assignment by the phishing simulation product”;
  • simulation messages reported with a third-party tool are not captured in Attack Simulation Training reports, and Attack Simulation Training does not target shared mailboxes.

With nophi.sh, there are two routes. The first is the nophi.sh report button for Outlook, deployed from the Microsoft 365 admin center, Settings, Integrated apps, with the manifest URL shown in the console. The second keeps Microsoft’s built-in button: it sends to your Exchange Online reporting mailbox, which forwards to the nophi.sh reporting address. In that case, check the outbound spam policy: since 2021, automatic external forwarding is off by default in most tenants (external forwarding control, August 18, 2026). A custom outbound policy scoped to that one mailbox allows it without opening it to everyone. Either way, the employee gets an AI verdict with its reasons, usually in one to two minutes.

Check that the policy applies (System override source = Phishing simulation)

Send a simulation to three or four mailboxes you control: a standard account, an executive account, a mailbox with MFA. The message must land in the inbox, not in quarantine. Then, in the Defender portal:

  • Threat Explorer (Plan 2) or Real-time detections (Plan 1): filter on System override source and select Phishing simulation. Your sends should be there;
  • email entity page: the Override(s) section shows Phishing simulation under Tenant override;
  • Threat protection status report: the System override view shows messages allowed by the policy;
  • advanced hunting: the value is in OrgLevelPolicy in the EmailEvents table.

If nothing shows up, the domain and IP match is wrong: go back to the list in the previous section. Keep the values of a successful send as a reference for the day something changes. With nophi.sh, the test email sent from the console does this check for you: if it reaches the inbox and you confirm it, delivery is proven.

Run the first campaign

Once delivery is verified, import your employees, with a Microsoft 365 (Entra ID) sync or a CSV file, and start with a small group before widening. A first campaign measures a funnel: open, click, then credentials typed on the fake page, where passwords are never collected. For a public benchmark, KnowBe4 reports an initial click rate of 33.1%, down to 4.1% after twelve months of training (KnowBe4, Phishing by Industry Benchmarking Report 2025, measured on its customers).

When an employee clicks, nophi.sh shows them the clues they missed right away, then assigns a short course on the theme of the trap, with reminders before the due date. Autopilot then retests each person every 7, 14 or 30 days depending on their risk level. nophi.sh costs €1 per person per month excl. VAT, minimum 10 people, no commitment, with a 14-day trial and no card; details on the pricing page.

If some mailboxes are elsewhere, read our Google Workspace allow-list guide. Moving off GoPhish? Our guide to migrate from GoPhish covers targets, templates and results.

Frequently asked questions

How do I allow a third-party phishing simulation in Microsoft 365?

With the Advanced Delivery policy in the Microsoft Defender portal, Phishing simulation tab: add at least one domain (the envelope sender MAIL FROM domain or the DKIM domain) and at least one sending IP of the tool. A message must match one domain and one IP on the list. Microsoft 365 accepts no other allow-list for messages classified as high confidence phishing.

Do I need Defender for Office 365 Plan 2 or E5 for Advanced Delivery?

No. According to Microsoft Learn, the policy is part of the built-in security features for all cloud mailboxes, and is also in Defender for Office 365 Plan 1 and Plan 2. Microsoft 365 Business Premium, which includes Plan 1, is enough. You need the Security Administrator role group plus Organization Management in Exchange Online, or an equivalent Entra role.

How many domains and IP addresses can I add?

In the portal: up to 50 domains, 10 sending IP entries (single IP, range or CIDR block, IPv4) and 30 simulation URLs. The reference page of the New-ExoPhishSimOverrideRule cmdlet says 20 domains at most and 10 IP entries. IPv6 can only be added in PowerShell.

How do I add an IPv6 address to Advanced Delivery?

In Exchange Online PowerShell, because the portal only takes IPv4. Create the policy with New-PhishSimOverridePolicy -Name PhishSimOverridePolicy, then the rule with New-ExoPhishSimOverrideRule -Policy PhishSimOverridePolicy -Domains your-domain -SenderIpRanges 85.208.144.52,2a09:8740:0:3::52. If the rule already exists: Get-ExoPhishSimOverrideRule | Set-ExoPhishSimOverrideRule -AddSenderIpRanges 2a09:8740:0:3::52.

Do I need to fill in the simulation URLs field?

Microsoft says this setting isn’t required for links in email phishing simulations: it is for links sent in Teams messages or Office documents. Filling it in does no harm. The nophi.sh console suggests app.nophi.sh/t/* if you want to set it.

Why are my simulation emails still quarantined as High Confidence Phish?

Most often the domain and IP match fails: the domain you added is neither the MAIL FROM domain nor the DKIM domain, or the IP Microsoft sees is not the one you added (relay, gateway in front of Microsoft 365). Microsoft also lists unsupported cases: simulations inside the organisation (DIR:INT), hybrid routing through Exchange Server, in-and-out routing through a third-party service, and Direct Injection.

Does the built-in Outlook Report button work in a shared mailbox?

Yes, in Outlook for Mac, iOS, Android, the new Outlook for Windows and Outlook on the web. The delegate needs Send As permission on the shared mailbox; without it, the message is only removed from the folder and does not reach the reporting mailbox.

How do I check that a message was delivered by the policy?

In Threat Explorer (Plan 2) or Real-time detections (Plan 1), filter on System override source and select Phishing simulation. On the email entity page, the Override(s) section shows Phishing simulation under Tenant override. In advanced hunting, the value is in OrgLevelPolicy in the EmailEvents table.