Synthetic Identities — The Fraud AI Didn't Invent, But Definitely Supercharged

This week’s inspiration comes from one of my favorite recurring conversations with Rajeev Gupta, who recently earned his certification in Anti-Money Laundering (AML). One term he kept circling back to: synthetic identities.

I’m someone who’s genuinely inspired by new technology, and excited about what AI keeps unlocking. But part of that curiosity is also being honest about where it cuts both ways. This is one of those places.

So what is a synthetic identity?

Unlike classic identity theft — where a fraudster steals your whole identity — a synthetic identity is a Frankenstein construction. Take a real Social Security number (often a child’s or someone who doesn’t actively monitor credit), combine it with a fabricated name, date of birth, and address, and you’ve built a “person” who doesn’t exist but who can open bank accounts, apply for credit, and slowly build a legitimate-looking financial history.

The slow build, then the exploit

Here’s what makes this fraud pattern so effective: patience. A synthetic identity is often cultivated for weeks, months, even years — making small purchases, paying bills on time, slowly earning credit limit increases, sometimes even getting added as an authorized user on someone else’s account to inherit good credit history. To the system, this “person” looks like a model customer, quietly aging into trustworthiness.

Then comes the exploit: the fraudster maxes out every line of credit at once, drains the accounts, and disappears — a move often called a “bust-out.” Because there was never a real victim watching a credit report, the fraud can go unnoticed until the bill collectors start calling an identity that never existed.

Why AI changes the math

This isn’t a new scam. But LLMs and generative AI have quietly removed most of the friction that used to limit it. Convincing fake photos, backstories, chat histories, even documents — the kind of supporting “evidence” a synthetic identity needs to look real during that slow build — used to take skill and time to fabricate. Now it takes a prompt. What used to be the domain of organized fraud rings is increasingly within reach of, as I’d put it, “nearly anybody who wants to.”

The pattern to watch

Like every technology shift, this one comes with a new category of risk that institutions — banks, insurers, HR departments — need to actively build defenses against, not just be aware of. AI didn’t invent synthetic identities. But it just gave the technique a serious upgrade in speed and accessibility.

Worth forming your own opinion on where you think the guardrails need to go.

Inspired by a wonderfull chat with a good friend Rajeev Gupta