McDonald's Exposed 64 Million Job Applicants. The Password Was "123456." Here's What That Means for Your Business.

Case Study AI Security

May 2026  ·  7 min read

Security researchers spent 30 minutes reviewing McDonald's AI hiring chatbot. They typed "123456" into the admin login. It worked. They now had access to 64 million job applications — names, addresses, phone numbers, chat transcripts, personality test results. This wasn't a sophisticated attack. It was the most predictable failure in security. And it's happening in small businesses every day.

---

Ian Carroll had been using McDonald's AI hiring chatbot — a system called McHire, built by Paradox.ai, featuring an AI recruiter named "Olivia" — when he noticed something odd about the platform's backend. He mentioned it to fellow researcher Sam Curry. Together they spent about 30 minutes looking around.

At some point one of them tried the most obvious thing possible. They typed "123456" as the username and "123456" as the password on an administrator login page.

They were immediately logged in.

"Without much thought, we entered '123456' as the password and were surprised to see we were immediately logged in," Carroll later wrote. From there they found a second vulnerability — an insecure API that let them pull any applicant's record by simply changing a number in a URL. They retrieved full applicant PII including chat transcripts, contact info, job-form data, timestamps, shift preferences, personality test outcomes, and tokens that could impersonate candidates.

Total time to access 64 million records: about 30 minutes. No malware. No exploit. No nation-state resources. Just a default password that nobody had ever changed.

64M job applications exposed — names, addresses, phone numbers, chat transcripts, personality test results Carroll & Curry / TechCrunch, June 2025

30 min how long security researchers spent reviewing the platform before finding full administrative access Ian Carroll researcher blog, 2025

187% growth in enterprise AI adoption 2023–2025, while security spending grew only 43% — a widening security deficit Oasis Security analysis, 2025

"This incident is a prime example of what happens when organizations deploy technology without an understanding of how it works or how it can be operated by untrusted users." — Evan Dornbush, CEO of Desired Effect, to CSO Online. The same observation applies to every AI tool deployed by every organization that moved fast and secured slowly.

What actually happened — and why it matters beyond McDonald's

The McHire platform was built by Paradox.ai and deployed by McDonald's to handle the first stage of job applications at scale. Olivia, the AI chatbot, conducted initial interviews, collected applicant data, and routed candidates through the hiring pipeline across years of McDonald's hiring activity.

The vulnerability wasn't in the AI. It wasn't a sophisticated attack. A test administrator account had been set up with default credentials — username "123456," password "123456" — and those credentials were never changed, never rotated, and never flagged as a risk. The account sat there, fully functional, connected to 64 million real applicant records, waiting for someone to try the obvious.

McHire Admin Login — mchire.paradox.ai

$attempting admin login...

username: 123456   password: 123456

✓ login successful — administrator access granted

$enumerating applicant records via IDOR vulnerability...

✓ record 1: name, email, phone, address, chat transcript

✓ record 2: name, email, phone, address, personality test results

... continuing across ~64,000,000 records ...

TOTAL TIME ELAPSED: ~30 minutes

The five lessons every small business needs to take from this

1

Every tool you deploy has an admin login. Do you know the password?

Default credentials

Every third-party tool, SaaS platform, AI application, or vendor-managed system your business uses has administrator credentials. Some were set by the vendor during setup. Some were set by an IT contractor who left two years ago. Some have never been changed from the default.

The question right now: can you name every system that has an admin login, confirm who has access, and confirm the password isn't "admin," "password," or "123456"? Most SMBs cannot answer this for every system they run.

Action this week Pull a list of every platform, tool, and vendor portal your business uses. For each one: confirm who the admin account belongs to, when the password was last changed, and whether the original setup credentials were ever rotated. This takes an afternoon. The McDonald's breach took 30 minutes to execute.

2

AI tools are not secured by default — they're deployed by default

AI security gap

Enterprise AI adoption grew 187% between 2023 and 2025. Security spending grew 43%. The gap between how fast AI tools are being deployed and how thoroughly they're being secured is the defining security problem of 2026. Every AI tool that touches customer, employee, or business data is a potential exposure point — and most were deployed to solve a business problem, not a security problem.

The 2026 Verizon DBIR documented that shadow AI tripled to 45% of employees. The McDonald's breach shows what happens when even approved, vendor-supplied AI tools aren't secured. Add unsanctioned tools to the picture and the exposure surface multiplies.

The question to ask about every AI tool Who has admin access? What data does it touch? Were default credentials changed? Is there MFA on the admin account? Was a security review done before deployment — or did it go live because it solved a problem and security was assumed?

3

Vendor security is your security — you're responsible for what you deploy

Third-party risk

Paradox.ai built the platform. McDonald's deployed it without auditing the admin credentials. The 64 million people whose data was exposed didn't know Paradox.ai existed — they applied for a job at McDonald's. From a legal and reputational standpoint, the exposed data belongs to McDonald's and the regulatory consequences fall on McDonald's.

The 2026 DBIR documented that third-party breaches now account for 48% of all incidents — up 60% in one year. You deploy a vendor tool, the vendor has a security gap, and your customers pay the price while your brand takes the damage.

For your business For every vendor tool that handles customer or employee data, ask before deployment: has this tool passed a security review, and do I understand who has admin access to my data within their platform? The answer doesn't need to be perfect. It needs to not be "we never thought about it."

4

The data you collect for one purpose becomes liability for another

Data minimization

McDonald's collected applicant data to fill jobs. That data — names, addresses, phone numbers, personality test results, chat transcripts — became a 64-million-record liability when the platform was misconfigured. Every piece of data your business collects and retains is a potential liability in a breach.

For your business A simple data audit — what do we have, where is it stored, who can access it, do we still need it — is one of the highest-value security exercises an SMB can do. Data you don't retain can't be breached. Data you retain but don't actively need is liability with no upside.

5

If it happened to McDonald's, it's happening somewhere in your stack right now

The uncomfortable truth

McDonald's has IT teams, security budgets, and vendor management processes. They still deployed an AI tool with a default password nobody changed. Not because they were careless — because the pace of AI deployment has outrun the pace of security review everywhere, at every scale, in every industry.

The uncomfortable question: how many tools in your stack were deployed to solve a problem, secured by assumption, and never audited? The McHire admin account sat open for years before anyone checked. What's been sitting open in yours?

The one action that changes this Schedule a 30-minute audit of every tool your business uses that handles customer or employee data. Confirm the admin credentials aren't default. Confirm MFA is on. Confirm who has access. The McDonald's breach took 30 minutes to execute. The audit that would have prevented it would have taken 15.

The exposure — names, email addresses, home addresses, phone numbers, chat transcripts, personality test results — is exactly the data attackers use to launch targeted phishing campaigns. A criminal who accesses 64 million job applicants doesn't need to crack a single password. They have enough information to send 64 million personalized, plausible phishing emails. The breach isn't just an exposure — it's a targeting database.

The pattern that connects this to every post in this series

The McDonald's breach is not a unique story. It is the same story this series has told from twelve different angles: the gap between the security tools organizations deploy and the security practices they actually maintain. A password policy without a password manager. MFA available but not enforced. A vendor tool deployed to solve a business problem, secured by assumption, and never audited.

The credential series, the clean scan post, the 30-minute audit, the CISA GitHub leak, the Verizon DBIR — they all describe versions of the same failure. The McHire admin account with "123456" is the most readable version because everyone understands what "123456" means. It means nobody checked. And in 2026, with AI tools being deployed faster than they're being secured, "nobody checked" is the most common security posture in business.

The Veriti Spottr CyberScore is built to surface exactly what nobody checked — the credential exposures, the configuration gaps, and the external attack surface signals that sit open until someone spends 30 minutes looking. The McDonald's breach was found by researchers with good intentions. The next one at your business will be found by someone who doesn't have them.

📚 Credential Security Series — Read the full series

Part 1 → Your Password Policy Isn't Protecting You. Your Employees' Habits Are. Part 2 → Your Credentials Are Already For Sale. Here's How to Find Out in 60 Seconds. Part 3 → You Turned MFA On. Attackers Already Have Three Ways Around It. Part 4 → What Attackers Do With Your Data in the First 60 Minutes Part 5 → The 30-Minute Security Audit Every SMB Owner Should Do This Week Part 6 → The First 24 Hours After You Discover a Breach

Find what nobody checked in your business. Veriti Spottr's beta is free.

Get your CyberScore →

VS

Veriti Spottr Team AI-powered cyber risk clarity for SMBs  ·  veritispottr.com