Here’s the short version: domain hijacking is someone else taking control of the domain name, usually through the registrar login or the DNS panel, so mail and the website follow their records. Lock the name at the registrar, put MFA on every login that can change DNS, and turn on auto-renew with a company payment method. Adding the domain in Microsoft 365 does not move ownership to Microsoft.
The site can look fine on a Tuesday and be someone else’s site on Wednesday, with the server in the same rack. The change was not on the server. It was in the record that tells the internet where the server is.
What domain hijacking actually changes
A domain name is a registration. The registrant is the party the name is held for. The registrar is the company that sponsors that registration. DNS is the set of records, including the nameservers, the website address, and the mail servers, that the world queries. Domain hijacking, in the cases that hurt a small firm, is a change to one of those controls by someone who should not have it.
Three changes matter more than the others:
- Nameservers move, and every record the firm thought it had stops being the record the world sees.
- The registrant contact or the account email changes, and the firm stops receiving the warnings.
- The name transfers to another registrar, and the old portal no longer has it.
The website files can be untouched through all three. Staff will say the site is “down.” It is not down. It is answering from somewhere else, or not answering because the new records point nowhere useful. Mail does the same thing. A password reset for the Microsoft 365 tenant starts going to whatever MX record is live.
CISA’s guide on mitigating DNS infrastructure tampering, written for organizations outside the federal directive, tells you to review the nameserver delegation and the records on the authoritative servers, and to treat a mismatch as a security incident. The same guide says to change passwords on every account that can edit DNS, including the registrar and any third-party DNS panel, and to put MFA on those accounts. If a provider cannot do MFA, CISA’s advice is to consider moving to one that can.
The registrar login is the asset

Whoever can sign in at the registrar can usually change nameservers, edit contacts, and turn the transfer lock off. That login is more valuable than the hosting password. The National Cyber Security Centre’s guidance on managing public domain names says the important control, for a firm with one public name, is 2-step verification at the registrar. It also says that if a web host registered the name, the organisation should manage the domain, not leave it in the host’s account.
We ask a plain question on the first call. Whose name is on the registrar account, and who can pass the second factor today? A designer who registered the name in 2019 as a favor, on a personal email, is a common answer in a small firm. That is not a moral failure by the designer. It is a single point of failure. If they leave, get locked out, or have that mailbox taken, the firm does not have its own name. The NCSC’s point is that the organisation should hold the registration. Two named people at the firm should be able to sign in. A shared “the registrar password” in a note is the same shape of problem as a shared local administrator password.
CISA’s original Emergency Directive 19-01, which the agency closed on January 8, 2026 after deciding the federal actions had been met, described DNS tampering that started with accounts able to change records. The directive told agencies to use factors that resist phishing, and it said SMS-based MFA is not recommended, consistent with NIST SP 800-63B. The directive no longer binds those agencies. The recommendation is still the right one for a registrar login: an authenticator app or a security key, not a text message to a phone the front desk shares. Our multi-factor authentication guide is the same preference on the tenant. The registrar is a separate vendor, with its own toggle.
Registrar lock is a status code, not a feeling
ICANN publishes the EPP status codes a registry actually enforces. The normal state, ok, means the name has no pending operation and no prohibition. ICANN’s note on that state is to ask the registrar for restrictions. The three that matter for a small firm are:
| Status | Who sets it | What it refuses |
|---|---|---|
| clientTransferProhibited | The registrar, from your account | A transfer to another registrar |
| clientDeleteProhibited | The registrar | Deletion of the name |
| clientUpdateProhibited | The registrar | Updates to the domain at the registry |
ICANN’s description of clientTransferProhibited is specific. It tells the registry to reject a transfer. That is the control aimed at hijacking and fraud. If you do want to move the name, you ask the registrar to remove the status first. Leave it on the rest of the year. An authorization code is how a planned move is confirmed. It is not a substitute for the lock. While the transfer-prohibited status is on, the registry is supposed to reject the transfer.
Client codes come off with the same login
The NCSC’s guidance for registrars on security features draws the line registrants need to understand. A client status is one the registrar’s own system can change. A server status is one the registry sets, and removing it is a separate process. Client locks are still worth having. They stop a transfer that only has an auth code, and they stop a casual delete. They do not survive a person who is already inside the registrar account and turns them off.
Registry lock is extra, where it exists
ICANN describes serverTransferProhibited as uncommon. It shows up in disputes, at the registrant’s request, or during a redemption period. Some registries sell a registry lock service, which a registrant reaches through the registrar. Taking that status off takes longer, because the registrar has to ask the registry. A 12-person firm should not stall the basics while shopping for registry lock. Get the client locks, MFA, and auto-renew in place. Ask whether registry lock is even offered for the extensions you hold. If it is, and the name is the one the firm cannot operate without, it is the stronger lock. If the registrar’s page does not mention it, do not invent a checkbox.
The DNS panel can hijack mail without a transfer
Many firms do not host DNS at the registrar. The registrar holds the nameservers, and those nameservers point at Cloudflare, at the web host, or at Microsoft. A person who can edit that second panel can change the website address and the MX records without transferring the domain and without touching clientTransferProhibited. CISA’s guide treats that panel as the same class of account as the registrar: password change, MFA, and a review that the records still point where you think.
This is why “the lock is on” is an incomplete sentence. The lock at the registrar can be perfect while the DNS host account is a password from 2017 with no second factor. List both logins. If they are the same vendor, say so. If they are two vendors, two people should be able to get into each, and a departure should remove the leaver from both. That removal belongs on the employee offboarding checklist, next to the mailbox and the laptop. A designer or a former office manager with a standing login is still a path in.

If the records have already changed, treat it as an incident, not as a hosting ticket. Call the registrar using a phone number from your own notes or from the vendor’s site you already bookmarked, not from a link in the alarming email. Our incident response plan is the wider sequence. The domain-specific part is short: who can still sign in, whether the lock is on, whether the nameservers match last month, and whether the account email was changed. Do not try to “take it back” by registering a lookalike name and emailing customers from that. That is a second incident.
Microsoft 365 does not hold the domain
Adding a custom domain in Microsoft 365 proves the tenant could see a DNS record you published. After that, Exchange and the admin center will talk about the name as if it lives there. It does not. The registration remains at the registrar. If you let Microsoft host DNS, Microsoft’s nameservers are a choice you can undo from the registrar. If you keep DNS elsewhere, Microsoft only has the records you copied in. Either way, a Global Admin can reset a user. A Global Admin cannot, by that role alone, stop a nameserver change at a registrar Microsoft does not operate.
When nameservers move, Microsoft 365 mail breaks even though the tenant is healthy. Password resets and encrypted messages follow the new MX. Microsoft 365 message encryption protects a message you send. It does not pin the domain to your registrar. The order of operations on a bad day is the name first, then the mail.
Auto-renew, and an email that is not only on this domain
The NCSC tells organisations to turn auto-renew on, because a domain has an expiry, after which administrative access is unavailable and another organisation can register the name. Pay with a method the organisation maintains. A personal card that expires, or a card tied to someone who left, is how a name lapses in a quiet month. Put the renewal date on a calendar a second person can see.
The warning email has to survive this domain
The contact email on the registration is how the registrar warns you about a transfer, a contact change, or a failed payment. If that mailbox is admin@ on the same domain, and the MX record is what just got changed, the warning goes to the person who made the change. Use a mailbox the firm controls that does not depend on this one name, or a second domain you also lock. Check the contact at least when someone who had the login leaves.
DMARC does not stop a nameserver change
DMARC is still worth publishing. It tells receiving mail servers what to do with messages that claim to be from your domain and fail authentication. It lives in DNS as a TXT record. Someone who can replace your DNS can replace that TXT, and can publish mail records of their own. DMARC assumes you still operate the domain. Domain hijacking is the loss of that assumption. Do both. Do not let a clean DMARC report convince you the registrar account is fine. The report only describes mail that tried to use the name. It does not describe who can edit the name.
What we check before we call the name locked
On a small firm’s primary domain, the list is short:
- The registrar account is in the firm’s name, with two people who can pass MFA, and the second factor is not SMS if the registrar offers something stronger.
- clientTransferProhibited is on. clientDeleteProhibited and clientUpdateProhibited are on unless a change is underway that day.
- The nameservers match the DNS host you think you pay. That host has its own MFA, and a leaver is not still a user.
- Auto-renew is on, the card is a company card, and the expiry is on a shared calendar.
- The registrant email is one the firm will still read if this domain’s mail breaks.
- Microsoft 365, the website host, and the registrar are written down as three logins, not one.
We will not call a privacy proxy a lock. Hiding the contact address does not reject a transfer. We will not call a DMARC pass a lock. We will not call a Microsoft 365 custom domain a registration. We will not leave the only set of credentials with a vendor who is not on a contract the firm can enforce when that vendor stops answering.
If the registrar login, the DNS host, and the renewal card should be put in the firm’s name as one pass, start with cybersecurity and contact Secure Techies. The website is replaceable. The name it answers to is the thing that is hard to get back.
