Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Happened in the MGM Resorts Attack?

In September 2023, a ten-minute phone call to an IT help desk led to ransomware across MGM Resorts, roughly ten days of disruption, and a reported impact of about 100 million US dollars.

Illustration for the cybersecurity term MGM Resorts Attack
Home / Blog / Cybersecurity Terms / What Happened in the MGM Resorts Attack?
17 November 2026

What Happened in the MGM Resorts Attack?

What Happened in the MGM Resorts Attack?

On 10 September 2023, MGM Resorts International began shutting down systems across its hotels and casinos. Guests queued for paper room keys, slot machines went dark, the booking website stopped taking reservations, and restaurant point-of-sale terminals reverted to cash. The disruption lasted around ten days. In its quarterly filing, MGM put the financial impact at approximately 100 million US dollars, plus about 10 million in one-off technology and legal costs.

What makes the case instructive is how little technical sophistication the entry required. The group responsible, tracked as Scattered Spider, UNC3944, or Octo Tempest, identified an MGM employee on LinkedIn, telephoned the IT service desk, impersonated that employee convincingly enough to have the account’s multi-factor authentication reset, and walked in.

From a Phone Call to Domain Control

Once inside, the attackers escalated to super administrator rights in MGM’s Okta tenant and gained control of its Azure environment. That combination is close to total: an identity provider administrator can create accounts, disable logging, add federation trusts, and mint access for themselves anywhere single sign-on reaches.

After several days of reconnaissance and data theft, the group deployed ALPHV ransomware, also known as BlackCat, against roughly a hundred ESXi hypervisors. Encrypting hypervisors rather than guest operating systems is an efficient tactic: a single host holds dozens of virtual servers, and endpoint agents generally do not run on the hypervisor itself.

The Caesars Comparison

The same group had reached Caesars Entertainment weeks earlier through an outsourced IT support provider. Caesars disclosed the incident in an 8-K filing and was widely reported to have paid roughly 15 million US dollars against an initial demand of about 30 million. Caesars avoided the visible operational collapse; MGM did not pay and absorbed the outage instead.

Neither outcome is a model. Caesars still suffered the loss of a loyalty programme database containing driving licence and social security numbers, and still faced the litigation and regulatory follow-up that attends any data breach. The comparison is useful mainly because it shows the ransom decision does not determine whether the organisation was harmed.

What Was Taken

In October 2023, MGM confirmed that personal data belonging to customers who had transacted with the group before March 2019 had been stolen. The categories included names, contact details, dates of birth, and driving licence numbers, with passport numbers and social security numbers affected for a smaller group. That dataset drove a consolidated class action and years of consumer notification work.

Why This Pattern Reached the UK

Scattered Spider is not a distant problem. In spring 2025, UK retailers including Marks and Spencer, Co-op, and Harrods were hit by intrusions attributed to the same tradecraft, with the NCSC issuing specific advice on help desk identity verification in response. Several of the individuals charged over the campaign have been British teenagers, and arrests were made in both the UK and the United States during 2024 and 2025.

The defensive gaps the group exploits are consistent:

Controls That Would Have Mattered

Verification at the help desk is the single highest-value change: require a callback to a registered number, a manager’s confirmation, or an in-person check with photographic identification before any factor is reset, and log every reset for review. Separate privileged identity administration into dedicated accounts on dedicated devices. Alert on the handful of Okta and Entra ID events that precede a takeover, such as new federation trusts, administrator role grants, and changes to logging. Harden ESXi by disabling SSH, enforcing lockdown mode, and keeping hypervisor management on an isolated network.

Testing those assumptions is exactly what Social Engineering Services are for; a controlled call to your own service desk usually answers the question faster than a policy review does.

FAQ – Frequently Asked Questions

Who Is Scattered Spider?

Scattered Spider is a loosely organised, largely English-speaking criminal group also tracked as UNC3944 and Octo Tempest. It specialises in social engineering against help desks and identity providers rather than exploiting software vulnerabilities, then deploys ransomware supplied by established affiliate programmes. Members have been arrested in the UK and the United States, but the tradecraft has spread far beyond the original group.

Did MGM Pay the Ransom?

MGM did not pay. It chose to shut down systems, rebuild, and absorb roughly ten days of operational disruption, reporting an impact of about 100 million US dollars. Caesars Entertainment, attacked by the same group shortly beforehand, reportedly paid around 15 million US dollars and avoided the visible outage, although its stolen loyalty database was still lost.

How Do You Stop a Help Desk Being Social Engineered?

Remove the discretion. Require a fixed verification procedure before any password or authentication factor is reset: a callback to the number held in the HR record, confirmation from a named line manager, or video verification against a stored photograph. Record every reset, review them weekly, and run periodic simulated calls so the process is tested rather than assumed.

Why Did the Attackers Target VMware ESXi?

Hypervisors concentrate value. Encrypting a hundred ESXi hosts takes down the hundreds or thousands of virtual servers running on them, and hypervisors rarely run the endpoint detection agents that would spot ransomware on a Windows server. Many estates also allow management access to ESXi from the general corporate network, so one set of stolen administrator credentials is enough.

Related Terms

You May Be Interested In These

What Is the DREAD Risk Scoring Model?

Cybersecurity Terms

17 November 2026

What Is the DREAD Risk Scoring Model?

Read More
What Is PetitPotam?

Cybersecurity Terms

17 November 2026

What Is PetitPotam?

Read More
What Are Loaders and Droppers?

Cybersecurity Terms

17 November 2026

What Are Loaders and Droppers?

Read More