When the Bank Hung Up on the Pope: What Identity Verification Really Verifies

A Chicago bank put Pope Leo on hold to verify his identity. He answered every security question correctly. Then, perhaps sensing their hesitation, he offered: "Would it help if I told you I'm Pope Leo?"

They hung up on him.

His only real crime? He'd moved to Rome and couldn't walk into the branch.

I've spent years implementing and auditing identity verification systems for financial institutions. This is the most honest demonstration I've ever seen of how these systems actually work in practice.

Because during that same period, an online age-verification tool was waving through children who'd drawn mustaches on their faces with a Sharpie marker.

Let's update the scoreboard:

• Third-grader with a marker — verified ✓

• Vicar of Christ, 1.4 billion followers, answered every security question — hung up on ✗

We Automated Away the One Thing That Worked

Remember when your local bank manager knew your face? Not your mother's maiden name or the street you grew up on — your actual face, attached to a reputation built over years of interactions.

That system had problems, sure. It didn't scale. It introduced bias. It limited your banking to business hours in one physical location.

So we replaced the bank manager who knew your face with a rules engine that knows your checkbox.

The new system can't tell a fraudster from a pontiff. It only knows whether you checked the right boxes and whether you walked into the right building.

This isn't a technology failure. The technology is working exactly as designed. The system successfully detected an anomaly: a customer whose behavior pattern suddenly changed. Geography shifted. Access method shifted. Risk score elevated. Protocol executed.

The Pope failed the system's real test — not "Are you who you say you are?" but "Are you behaving the way we expect you to behave?"

The Defense Nobody Wants to Hear

In fairness, even the Pope's own friend defended the bank's decision: "If someone calls me and tells me they're the Pope, I'm hanging up too."

Hard to argue with that logic.

This is the uncomfortable middle ground where I spend most of my time advising clients. A system that rejects the Pope and approves a Sharpie mustache isn't verifying identity. It's verifying compliance.

The bank followed its protocol perfectly. The age-verification system followed its protocol perfectly. One protocol said "anomalous behavior pattern = reject." The other said "image contains face-like features = approve."

Both systems optimized for the metric they were given. Neither was given "actually verify this is the right person."

The Pattern We Keep Repeating

I watched this same movie play out with airport security after 9/11. We built elaborate systems to catch the previous attack — shoe bombs, liquid explosives, belt buckles. The theater of security got more sophisticated while the actual security got more brittle.

We perfected catching threats that matched our checklist. We got worse at catching threats that didn't.

The TSA agent confiscating your water bottle isn't keeping you safer. They're demonstrating compliance with a protocol designed after someone else tried something different.

Your bank's identity verification system is the same architecture. It's optimized to catch the fraud patterns from 2018, executed through the compliance framework from 2022, defending against the threats we've already named and categorized.

Then AI Walks Into the Room

Now add AI voice clones to this equation.

I'm not talking about some future scenario. Right now, today, commercially available AI tools can clone a voice from a 3-second audio sample. They can pass voice biometric systems. They can navigate phone trees, answer security questions with information scraped from data breaches, and modulate tone to sound appropriately frustrated or cooperative.

These AI systems will sail through the same identity checks that stonewalled the Pope.

Why? Because they're optimized for the test. They know the checkbox. They exhibit expected behavior patterns. They don't trigger the anomaly detectors because they're designed specifically to color inside the lines.

The Pope triggered alerts because he was genuinely unusual. The AI voice clone won't trigger alerts because it's studied exactly what "usual" looks like.

A third-grader with a Sharpie gets through because the system checks "face detected" not "face is real." An AI voice gets through because the system checks "voice patterns match" not "human is real."

What We're Actually Measuring

I was reviewing an identity verification vendor's marketing materials last month. Impressive stats: 99.7% accuracy, sub-second verification times, machine learning models trained on millions of transactions.

I asked them one question: "When you say 99.7% accurate, what are you measuring? That the person is who they claim to be, or that the person successfully completed your verification steps?"

Long pause.

We've built increasingly sophisticated systems to verify that people can verify themselves. That's a different problem than verifying they are themselves.

The bank that hung up on the Pope had a perfect compliance record. Every checkbox checked. Every protocol followed. Every audit passed. They could demonstrate to regulators exactly why they rejected that caller — anomalous behavior, couldn't verify in person, claimed to be someone implausible.

They just couldn't demonstrate they'd actually protected anything.

The Control That Never Scaled

"Just come into a branch" used to be the fallback. The final line of defense. When the remote verification systems couldn't resolve your identity, physical presence solved it.

Except the Pope couldn't come into a branch. He was in Rome. Running the Vatican. Moderately busy schedule.

And your customers increasingly can't either. They're digital-native, mobile-first, branch-averse. The average age of someone who regularly visits a physical bank branch is 58 and climbing.

"Just come into a branch" is becoming "just prove you're not the kind of person who uses modern banking."

So what happens when the AI voice clone calls your bank, exhibits perfectly normal behavior patterns, has all the right answers scraped from the last three data breaches, and never triggers a single anomaly detector?

Your bank will wave them through. Because your bank's system is optimized to say yes to compliance, not no to sophisticated fraud.

The Question You Need to Ask Monday Morning

I'm not arguing we should go back to the bank manager who knew your face. That system's failures were real — bias, limited access, inability to scale, dependency on human judgment and memory.

But I am arguing we should be honest about what we replaced it with.

We didn't replace human judgment with better judgment. We replaced human judgment with consistent execution of rules. The rules work great until someone who doesn't match the rules shows up — whether that's the Pope or an AI that's studied exactly which rules to match.

Here's what I'm asking my clients to pressure-test right now:

Walk through your identity verification flow. Not the vendor's marketing deck — the actual implementation. Then ask your security team:

"If an AI voice clone called us today with data from the last breach and no unusual behavior patterns, what would stop us from waving them through?"

If the answer is "We'd catch them at step X" — great. Test it. Red team it. Actually verify your verification.

If the answer is "They'd have to come into a branch" — you've just made the Pope's problem your standard operating procedure.

And if the answer is a long pause, like my vendor got last month, you're verifying compliance, not identity.

The Pope would understand. He answered every question right and still got hung up on.

But the Sharpie mustache sailed through.

What's your system actually checking for?

Jay Schulman helps financial services firms navigate the gap between security theater and actual security. If your identity verification strategy relies on "nobody would actually try that" — we should talk.