AI Voice Cloning Scams Have Arrived — What Queens Businesses Need to Do Now

Your CFO calls, sounding exactly like your CFO, asking you to wire funds immediately for a time-sensitive deal. Except it isn't your CFO. It's an AI-generated clone of their voice, built from a few seconds of audio pulled off a webinar recording, a voicemail greeting, or a LinkedIn video.

This isn't speculative. In January 2024, the engineering firm Arup lost $25.6 million after a finance employee joined a video call where every other participant — including the CFO — was a deepfake. The FBI's Internet Crime Complaint Center logged its first-ever standalone AI-fraud category in its 2025 report: over 22,000 complaints and roughly $893 million in confirmed losses, and that figure is almost certainly an undercount, since most victims never realize AI was involved in the first place. Security researchers at Mandiant found voice phishing was the single most common way attackers broke into cloud systems in 2025 — ahead of email phishing.

Here's what's actually changed, and the verification protocol that neutralizes it regardless of how convincing the fake sounds.

What's Actually Different From Old-School Phone Scams

Caller ID spoofing and social engineering aren't new — Queens businesses have fielded fake "IT support" and "vendor" calls for years. What's new is that the voice itself is no longer a reliable signal. Your team has spent years training on "does this sound urgent or off?" — and attackers have spent that same time training AI to sound exactly right.

That shift matters most for smaller organizations, not less. This was never just a large-enterprise problem; it's a problem for any business where one employee can pick up a phone and ask another to move money, share credentials, or grant access.

A Verification Protocol That Doesn't Depend on Trusting a Voice

The fix isn't teaching people to "listen harder." It's removing voice as the sole authorization method for anything that moves money, data, or access. Four changes do most of the work:

• A callback rule, not a call-forward rule. Any request to move funds or change payment details — even from someone who sounds exactly like your CEO — gets verified by calling that person back on a known number, never a number provided in the same call or email.
• A shared verification phrase for your finance team, changed periodically, that isn't something anyone would say in casual conversation or post online.
• A two-person rule for wire transfers and vendor payment changes above a set threshold, with the second approver reached through a separate communication channel.
• Slow down urgency instead of matching it. "This has to happen right now, don't tell anyone" is the oldest trick in the book, and it works exactly as well on a real-sounding voice as a cloned one.

What It Actually Costs When This Goes Wrong

The Arup case is the clearest example because it's so well documented: $25.6 million moved on the strength of a video call where the deepfakes were convincing enough to fool employees who'd worked with the real executives for years. But the FBI's $893 million figure — spread across more than 22,000 reported incidents in a single year — shows this isn't a one-off. It's a routine attack vector now, and the tooling to run it has gotten cheap enough that it doesn't take a sophisticated attacker to try it on a Queens business instead of a multinational.

Where This Intersects With What You're Already Required to Do

If you're in a regulated vertical, building this kind of verification isn't just good practice — it likely already falls under obligations you have. Financial firms under NY DFS 23 NYCRR 500 need to account for social-engineering vectors in their risk assessments, not just technical vulnerabilities. Law firms handling client funds face the same wire-verification exposure we've written about before with business email compromise — a cloned voice is simply a more convincing version of the same attack. Medical practices moving vendor or payroll payments carry the same risk with, frankly, less attention paid to it than the other two.