CNBC exec almost lost his crypto in minutes and was saved by an AI chatbot. Here's how to spot impersonation scams, even when they seem legitimate

What saved him was having to suspend the call to attend a work meeting. With the convincing voice no longer pressuring him to act, Gewirtz was able to take a beat and recognize a few red flags from the call. He even used artificial intelligence, as well as a contact of his who used to work at Coinbase, to figure out what was going on.

While Gewirtz was able to duck the scam, not every victim is so lucky. In fact, Gewirtz’s close call is a reminder that scams are becoming more sophisticated, and that if a veteran news executive at a major network nearly got tricked, just about anyone can.

How the scam almost worked

Like most people, Gewirtz doesn’t usually pick up the phone when it displays an unknown number. But since this particular call showed a San Francisco number, the location of Silicon Valley and one of CNBC’s key bureaus, Gewirtz decided to answer.

The man on the phone, who introduced himself as Brian Miller from Coinbase’s security office, told Gewirtz that his account had been targeted by suspicious login attempts from Frankfurt, Germany. The man framed the call as a protective measure, saying the goal was to stop a scammer before any damage occurred.

To make the story more convincing, the man leaned on key details in order to disarm Gewirtz’s skepticism. Spoofing the call with a 650 area code was the opening strategy; other techniques included referencing sensitive personal data and sending emails that appeared to come from official Coinbase addresses.

Playing psychological games was also key. The caller repeatedly emphasized urgency, warning that a scammer in Germany was actively trying to access Gewirtz’s account to make transfers.

“He’s trying to make another transfer,” the man on the phone told Gewirtz, “but I have it on hold so he can’t.”

The fear of losing access to his crypto kept Gewirtz on the line, even when some details — such as the mention of a photo in Gewirtz’s account that he was sure he never uploaded — didn’t add up.

It wasn’t until he got off the phone and ran the addresses from the emails the caller had sent through Claude — an AI chatbot from Anthropic — that Gewirtz realized he could be getting scammed. When Claude flagged the email addresses as “almost certainly a PHISHING scam,” Gewirtz quickly realized he had just dodged a bullet.

“I said to myself, ‘thanks, Claude,’ while also thinking, ‘that was close.’”

How to protect yourself as scamming efforts increase

Unfortunately, Gewirtz’s close call isn’t all that unusual, as it reflects a sharp rise in crypto scam attempts that are becoming harder to spot and easier to believe. According to Chainalysis, impersonation scams targeting cryptocurrency users surged by more than 1,400% year over year in 2025, with losses estimated at $17 billion (2).

Using AI to enhance their abilities, these scammers tend to create fear and urgency to prey on their victims. And as Gewirtz’s story has shown, this playbook has the potential to fool even experienced professionals.