In late September 2026, a few months after the Knicks beat the San Antonio Spurs for their first title since 1973, some Madison Square Garden employees received an email with the subject line “your Employee Championship Ring is here” [1][2]. The email offered a choice of two designs, “Championship Gold” or “Garden Classic,” at “no cost to you” [2][3]. Employees who clicked were told they had failed a phishing test and were sent to a mandatory cybersecurity class, even though some of them weren’t getting a real ring [1][2].
MSG later called the lure “an unfortunate mistake” and said it had been “recently hacked and extorted using an employee login and password obtained through a similar phishing attack,” which explains why it was testing staff in the first place [4]. Most of the public reaction focused on whether it was fair to use a ring as bait for the people who helped win it. For a security team, the more useful question is what a test like this can tell a company about its employees, and two of the largest phishing studies published so far give a fairly clear answer.
What the two studies found
Researchers at UC San Diego and ETH Zurich each ran simulated phishing campaigns against thousands of employees for most of a year or longer, and both groups presented their results at the IEEE Symposium on Security and Privacy [5][6]. The table below summarizes the main findings, and the sections after it go through them in more detail.
| UC San Diego Health study | ETH Zurich study | |
|---|---|---|
| Employees tracked | 19,789 | 14,733 |
| Length | 8 months (2023) | 15 months (2019 to 2020) |
| Simulated campaigns | 10 lures | 8 simulations per person |
| Annual security training | No significant effect on failure rates | Not tested |
| Training shown after a click | About 1.7 points lower failure rate on average | Trained employees clicked more often |
| Employees who failed at least once | 56% | 32% clicked, 25% took a dangerous action |
| What helped most | Not tested | Warning banners and a report button |
How much the lure affected click rates
The UC San Diego team sent ten different lures to the same group of employees, and the results varied widely depending on the email. Only 1.82 percent of employees fell for a fake Outlook password reset, while 30.8 percent fell for a fake update to the vacation and sick time policy, roughly 17 times as many [5]. A fake dress code update came close behind at 27.65 percent. The lures that worked best dealt with something employees cared about personally, and the ones that worked worst resembled the security warnings people have learned to distrust.
A free championship ring sent to Knicks staff after a 53-year wait falls well toward the personal end of that range. MSG’s click rate would therefore mostly reflect how appealing the ring was, and it couldn’t separate careless employees from people who simply received an unusually tempting email.
This also affects how a company should read its own results. Comparing a department that received an easy lure with one that received a hard lure doesn’t show which department is more careful, and a quarter in which the click rate fell may only mean the security team chose a less convincing email. To track how people improve over time, send the same lure, or lures of known difficulty, to every group you plan to compare, and report each result alongside the lure that produced it.
What the research says about training after a click
MSG used one of the most common designs in corporate phishing programs, in which an employee who clicks sees a page saying they failed and is then assigned a course. Both studies looked closely at how well that approach works.
At UC San Diego, employees who had recently completed annual security training failed at about the same rate as employees who hadn’t [5]. Embedded training, meaning the page shown right after a failed test, lowered the odds of failing by about 9.5 percent, which worked out to roughly 1.7 percentage points on average. More than half of those training sessions ended within 10 seconds, and only about a quarter were formally completed.
ETH Zurich found that the same design could make things worse. Employees who were shown voluntary training material after clicking went on to click more often than employees who weren’t, and 43 percent of those who remembered the training page said that seeing it made them feel safe [6]. The authors concluded that the method “may cause unexpected and negative side effects, such as increased susceptibility to phishing.”
Neither study claims that all training is useless. Both suggest that the low-effort, punitive version many programs run, a failure page followed by a mandatory module, achieves little and costs a good deal of employee goodwill. NIST’s guidance on learning programs already says phishing exercises “should not be punitive, nor should any employee be called out for their response,” and it recommends involving the legal team in the design of each exercise [7].
Why reporting is a better measure than click rate
The ETH Zurich researchers also gave employees a button for reporting suspicious emails and tracked how it was used. Staff filed 18,476 reports during the study, and about one in ten arrived within five minutes [6]. In the final five months, employees reported 918 real phishing emails connected to 252 large campaigns, and reviewing all the reports took human analysts roughly 1.5 emails a day. That’s a small workload for a detection network made up of thousands of employees. People who received positive feedback after reporting went on to report more often, and short warning banners on suspicious messages also reduced clicks considerably.
The UK’s National Cyber Security Centre makes a similar case. Its guidance tells organizations to measure “how many people reported it” alongside clicks, and it warns that “employees who are afraid for their jobs will not report mistakes.” It also advises checking with HR before running any phishing simulation [8]. Under MSG’s approach, an employee who clicked the ring email and then reported it would still have been sent to the class, which gives people little reason to report quickly the next time.
The following measures give a board a more accurate picture than click rate does.
| Metric | What it tells you |
|---|---|
| Time to first report | Roughly how long a real campaign would run before your team hears about it |
| Report rate | How many employees act as an early warning system, tracked separately for each lure |
| Credential submission rate | How many people would actually hand over a password, which matters far more than a click |
| Real phishing reported | Whether the reporting habit carries over to live attacks |
| Repeat credential submissions | Who needs a private conversation, kept out of team-wide reports |
Phishing attacks that don’t use email
According to MSG, its breach began with a stolen login and password [4]. That specific problem has a technical solution that doesn’t depend on employee judgment. CISA describes FIDO/WebAuthn as “the only widely available phishing-resistant authentication” [9], because a passkey or hardware security key won’t give its credential to a fake sign-in page. Accounts that can export customer data in bulk are the sensible place to start.
The group that claimed MSG’s June data leak used the name ShinyHunters, and Puck reported that owner James Dolan refused to pay the ransom [10]. The ShinyHunters name has also been linked to a technique that an email test can’t measure. In June 2025, Google’s Threat Intelligence Group described attackers who called employees while posing as IT support and talked them into approving a modified version of Salesforce’s Data Loader app, which the attackers then used to export data [11]. Google wrote that “in all observed cases, attackers relied on manipulating end users, not exploiting any vulnerability inherent to Salesforce.” MSG hasn’t said how its attackers got in, and more than one group has used the ShinyHunters name, but the technique is worth planning for because the employee never clicks a link.
Google’s recommendations for that campaign apply well beyond Salesforce. It advises limiting the ability to approve or manage connected apps to a small group of administrators and allowlisting the apps you know and trust. It also recommends restricting API access to people whose jobs require it and setting alerts on large data downloads, which Google calls “a common sign of Data Loader abuse” [11]. In Microsoft Entra ID and Google Workspace, the equivalent step is limiting which apps employees can grant access to on their own.
These consent screens now include a newer kind of app as well. Employees can connect AI assistants and agents to their mailboxes and CRM in a few clicks, and each approval gives that app ongoing access to company data that no phishing test will ever examine. Anzenna was built to cover this part of the problem. It connects to your identity provider and collaboration systems, and a browser extension adds visibility into AI tools, so it can inventory every OAuth-connected app and AI tool in use and tie each approval to the person who granted it. When a phished account or a newly approved app starts pulling data it has never touched before, Anzenna flags the activity, and sensitive data can be redacted or blocked as it moves. You can explore the console in our interactive tour, which runs on sample data, or book a walkthrough to see which apps hold the most access in your own environment.
How the ring email could have been used as a test
The ring email would have worked well as a test of reporting speed. MSG could have sent it to everyone without a failure page and measured how many minutes passed before the first employee reported it. Employees who reported it would get a thank-you, and employees who clicked would get a short explanation of what gave the email away, with the reporting time going on the board slide. That approach would have shown MSG how quickly its staff recognize the kind of personal, timely lure a real attacker would write, without sending anyone to a class over a ring they weren’t receiving. MSG says the rest of its staff will receive their real rings in about four to six weeks [3].
---
References
- The Athletic, “Knicks ring night, assistant coaches and MSG employees,” October 22, 2026.
- ESPN, “Phishing email to Knicks employees an ‘unfortunate mistake,’ MSG rep says,” October 23, 2026.
- NBC New York, “Knicks employees offered championship rings in phishing test email,” October 23, 2026.
- Associated Press via NY1, “Knicks say using championship rings in an email to fool employees was an ‘unfortunate mistake’,” October 23, 2026.
- Grant Ho, Ariana Mirian, Elisa Luo and others, “Understanding the Efficacy of Phishing Training in Practice,” IEEE Symposium on Security and Privacy, 2025.
- Daniele Lain, Kari Kostiainen and Srdjan Čapkun, “Phishing in Organizations: Findings from a Large-Scale and Long-Term Study,” IEEE Symposium on Security and Privacy, 2022.
- NIST, “SP 800-50r1: Building a Cybersecurity and Privacy Learning Program,” Section 2.7.2, September 2024.
- UK National Cyber Security Centre, “Phishing attacks: defending your organisation.”
- CISA, “Implementing Phishing-Resistant MFA,” October 2022.
- Puck, “The MSG hack that got buried by the Knicks parade,” June 22, 2026.
- Google Threat Intelligence Group, “The Cost of a Call: From Voice Phishing to Data Extortion,” June 4, 2025.
