123 Reg’s “AI Cyberattack”: What Was Taken, What Wasn’t, and What to Do

In brief

• 123 Reg says a September 18 AI-driven attack read customer contact and product data. No passwords were listed — but the phishing risk is real.
• The UK registrar, owned by GoDaddy, told customers in the week of October 26, 2026 that an intruder read contact information, related customer relationship data, company VAT numbers, and product information pulled through an API.
• The exposed fields — names, phone numbers, countries, language and currency preferences, account creation dates, products held, and marketing preferences — do not include passwords or payment details in the notice as reported. That limits the damage, but it hands an impersonator almost everything needed for a convincing fake renewal or support message.
• 123 Reg calls this “the first known AI-driven cyberattack against our company” and suspects a China-nexus group, citing attacker infrastructure in China and Singapore, AI “skills” found on Chinese servers, and a link to the ARTEX campaign that hit South Korean financial institutions. That attribution is the company’s own assessment, not a confirmed finding backed by a published technical report.
• If you are a 123 Reg customer: turn on two-step verification, and never trust a message claiming to be from 123 Reg — go to the website directly and log in there instead.

Last verified: October 29, 2026.

What happened

UK domain registrar 123 Reg told customers this week that it was the victim of what it calls an “AI-driven cyberattack.” The intrusion happened on September 18, 2026. Customers began receiving notification emails in the week of October 26 — roughly five weeks later — and the company disclosed the incident to The Register on October 28.

123 Reg described the September 18 intrusion as “the first known AI-driven cyberattack against our company.” That phrasing deserves a pause, because it is doing more work than it first appears to. It sounds like a claim about a new kind of attack. Read carefully, it is really three separate claims bundled into one label: that an intrusion happened, that AI tooling was involved in it, and that the attacker belongs to a particular group. Each of those claims stands on different evidence, and they should not all be believed with the same confidence. This article keeps them apart.

The company reported the attack to law enforcement agencies, brought in outside cybersecurity support, and notified the US Federal Trade Commission as well as data protection authorities in the UK and EU. It also said it worked with Singapore-based internet providers to shut down the attacker’s servers in the region, which helped disrupt the attack, alongside what it called “other actions by us and industry partners.”

What was taken, and what was not

The customer email said the compromised data included contact information, “related customer relationship data,” company VAT numbers, and “product information from an application programming interface (API).” A company spokesperson elaborated to The Register, and the field list is specific:

• Names and phone numbers
• Countries, language and currency preferences
• Account creation dates
• The products on each account
• Marketing preferences
• Company VAT numbers for business customers

Two things about that list matter. First, what is not on it: the email did not mention passwords or payment details being taken. That is a meaningful limit. Nobody is suggesting, on the current evidence, that an attacker walked away with the credentials to log in as you or with your card details.

Second, what is on it is more useful to a criminal than it looks. A scammer who knows your name, your phone number, the exact products on your account, when you created it, and your business VAT number does not need your password to sound legitimate. They can write a renewal reminder that names the right domain, a support message that references the right product, or a call that opens with details only your registrar should know. The exposed data is, in practical terms, a kit for building convincing impersonation. That is why the company’s own advice focuses on phishing, not on password resets.

There is one more wrinkle in the disclosure, and 123 Reg stated it openly. The customer email said the attacker’s activity “was limited to reading” customer data. But the spokesperson told The Register that “it’s possible they created copies of the information,” and that this possibility is what prompted the notification. Read that as honesty about uncertainty: the company can confirm the data was accessed, and it cannot rule out that copies were kept. Customers should act on the assumption that the data is out there, not on the hope that it was only glanced at.

No customer count was given in the disclosure coverage this draft is based on. Treat any number you see elsewhere with suspicion unless it comes with a named source.

What “AI-driven” actually means here

This is the part of the story most worth slowing down for, because “AI cyberattack” is becoming a label that explains less the more it is used.

What 123 Reg actually reported finding is narrower than the label suggests. Its investigators identified IP addresses they assessed to be core attacker infrastructure, hosted in China and Singapore. They also found AI cyberattack “skills” hosted on Chinese infrastructure, including copies of skills open sourced by Anthropic and several other AI exploitation frameworks and scripts. In other words, the attackers appear to have used AI-assisted tooling — agent-style workflows and pre-packaged attack “skills” — as part of the operation.

That is a real and important observation. It is not the same as an attack that was autonomous, or one that AI made possible in a way no human team could have achieved. Nothing in the disclosure shows a model acting on its own, or a capability that conventional attackers lack. The evidence supports “AI-enabled or agentic tools were used during the operation.” It does not prove that artificial intelligence ran the attack by itself, and it does not, on its own, identify who ran it.

Why does the distinction matter? Because it decides what defenders should do next. If the threat were a fully autonomous attacker, the answer would be some new class of defence. If the threat is a familiar intrusion carried out with AI-assisted tooling — faster probing, automated exploitation workflows, agents working through target lists — then the answer is the unglamorous work that already matters: close exposed APIs, watch for automated probing, and assume the attacker can move faster than a human team would. The 123 Reg case, on the evidence published so far, belongs in the second category.

The attribution case, and its limits

123 Reg told The Register that early signs point toward a China-nexus attacker, and that many of the artefacts left behind led it to suspect a Chinese state-sponsored group known for abusing Claude Code in attacks against at least 30 organizations. It also suspects the same group that hit a number of South Korea’s financial institutions using ARTEX, an open-source Chinese-developed agentic penetration-testing tool released in July 2026, was behind the attack on 123 Reg too.

The spokesperson pointed to three threads tying those claims together. First, the infrastructure finding: core attacker IP addresses hosted in China and Singapore, which the company said indicates “a strong Chinese nexus.” Second, the preparation artefacts: a mirror of the “anthropic_skills” repository and account naming conventions found on the Chinese servers, which the spokesperson said are consistent with preparation of this class of AI-augmented operation. Third, the timing overlap: on October 7, 2026, CrowdStrike published advisory CSA-261162, describing an unknown, likely Chinese-speaking actor that targeted several South Korean financial organizations starting in late September 2026 — the same period during which 123 Reg was attacked. That actor used ARTEX, and defenders observed ARTEX in the attack on 123 Reg as well.

That is a coherent case, and it is worth reporting. It is not a confirmed attribution. It is the company’s own assessment, relayed through a single publication, with no published technical report behind it. The group is unnamed. The specific ARTEX observations in the 123 Reg intrusion are described, not shown. The South Korea connection rests on a shared tool and an overlapping window, and shared tools are weak attribution evidence on their own — ARTEX is open source, and anyone can use it.

None of that means the assessment is wrong. It means readers should hold it at the right level of confidence: a credible lead from the victim’s investigators, not a verdict. If a technical report or a vendor advisory later confirms the link, the story upgrades. Until then, “123 Reg suspects” is the accurate verb, and any version of this story that drops the “suspects” is overstating the evidence.

Why a registrar breach is a phishing breach

It is tempting to file this story under “another company got hacked” and move on. For 123 Reg customers, the specific risk is narrower and more personal than that.

Domain registrars sit in a position of trust that most companies do not. Their messages arrive with renewal deadlines, transfer authorizations, and account warnings — messages customers are trained to act on quickly, because letting a domain lapse has real consequences. An attacker who knows your products, your account date, and your VAT number can imitate exactly those messages, and the imitation will survive a casual glance.

That is why 123 Reg’s own advice is worth quoting directly. The company told customers: “If you receive a message or call and aren’t sure it’s from 123 Reg, go directly to our website and log in.” And: “We will never contact you to request your password, payment details, or verification codes by email, text, or phone.”

Those two sentences are the whole defence in miniature. The first breaks the attacker’s channel: a link or a phone number in a suspicious message is never the way in, the website you navigate to yourself is. The second gives you a test for any future message: a request for a password, payment details, or a verification code, supposedly from 123 Reg, is by the company’s own statement not from 123 Reg at all.

What to do if you are a 123 Reg customer

Turn on two-step verification now. 123 Reg itself “strongly recommends” this, and the reasoning is straightforward: two-step verification protects your account even if someone tricks you into sharing your password. Find the setting in your 123 Reg account and enable it before you do anything else.

Treat every unexpected 123 Reg message as suspect until proven otherwise. If a message asks you to renew, confirm details, or follow a link, do not use the link or the number in the message. Go to the 123 Reg website directly, log in, and check your account status there. The same applies to phone calls: hang up and call back through the number on the website.

Watch for renewal and support impersonation in particular. The exposed fields — products held, account dates, business details — are exactly what a fake renewal notice or a fake support call needs to sound real. A message that names your domain correctly is not thereby legitimate. The test is the channel, not the content.

Do not panic about passwords or payments on the current evidence. The notice as reported does not list passwords or payment details among the exposed data. That could change if 123 Reg revises its notice, and this article will be updated if it does — but there is no basis in the current disclosure for assuming your login or your card details were taken. Changing your password is never harmful, but the highest-value action here is two-step verification plus the verify-through-the-website habit.

If you run a business on a 123 Reg account, brief anyone who handles the domain. The VAT number and product details in the exposed set mean a scammer can plausibly target an accounts or operations contact, not just the account holder. Make sure the people who might receive a renewal or transfer message know the rule: verify through the website, never through the message.

The wider pattern

The 123 Reg intrusion does not stand alone. It sits inside a month in which AI-assisted tooling has moved from a talking point to a recurring feature of intrusion reports: ARTEX in the South Korea financial attacks, AI agents in the ConnectWise investigation, and now AI “skills” found on infrastructure behind a registrar breach. The pattern is not that AI has replaced human attackers. It is that the tooling which lets a small team probe many targets, automate exploitation workflows, and move quickly once inside is now cheap, open source, and widely available.

That has a practical consequence for everyone else. The defences that matter against this class of attack are not exotic. They are the ones this story keeps returning to: do not leave APIs and management interfaces exposed to the internet, watch for automated probing, keep credentials out of shared workspaces, and build verification habits — like navigating to the website yourself — that survive even when the attacker knows your account details.

123 Reg’s disclosure is useful precisely because it is limited. It tells customers what was read, it admits what might have been copied, and it labels its attribution as suspicion rather than fact. The customers who take the practical half of that disclosure seriously — two-step verification on, suspicious messages verified through the website — will be better protected than customers of the next registrar breach who only remember the “AI cyberattack” headline.

Related reading

• ARTEX: The AI Pentesting Tool Used in the ConnectWise Hack
• South Korea AI Hacking Arrests: 90 Firms, 8.2M Records Stolen
• Law Firms Are the New ATM for Hackers, CrowdStrike Warns

Sources

• The Register, “123 Reg discloses ‘AI cyberattack’ on customer data, suspects China nexus,” October 28, 2026 — https://www.theregister.com/cyber-crime/2026/10/28/123-reg-discloses-ai-cyberattack-on-customer-data-suspects-china-nexus/5305936 (read in full, October 29, 2026; all intrusion, exposure, attribution, and response facts above come from this report unless noted)
• Let’s Data Science, “123 Reg discloses AI-driven cyberattack exposing customer data,” October 28, 2026 — https://letsdatascience.com/news/123-reg-discloses-ai-driven-cyberattack-exposing-customer-da-2a08ba28 (read in full, October 29, 2026; corroborates the field list, the “limited to reading” versus possible-copies distinction, and that the email did not mention passwords or payment details)
• CrowdStrike advisory CSA-261162, October 7, 2026 (unknown, likely Chinese-speaking actor targeting South Korean financial organizations; ARTEX use) — details as reported by The Register above; no standalone CrowdStrike advisory link is claimed here