Connecticut Lost $16.1 Million to AI Scams in 2025

The Scam Capital Nobody Wanted to Be

Connecticut has about 3.6 million people, so losing $16.1 million in a single year to AI scams means a lot of families got hurt. These weren't corporate breaches or stolen databases. They were phone calls. A voice on the line claiming to be someone you love, or a number on your screen that looked like your bank.

AI made those calls convincing in ways that weren't possible even two years ago.

According to CT News Junkie's analysis, scammers used three main tactics in 2025: voice cloning to impersonate relatives, spoofed caller ID to fake bank phone numbers, and deepfake celebrity videos to push fraudulent investment schemes. None of these tactics are new. The AI just made them work on people who never would have fallen for the old versions.

How Voice Cloning Actually Works

Cloning a voice doesn't require much audio. Researchers at Carnegie Mellon University have demonstrated convincing voice clones built from less than 30 seconds of source material, such as a voicemail, a birthday video on Facebook, or a clip from a local news segment.

Scammers clone the voice, then call a parent or grandparent in a moment of staged panic. The script usually involves an accident, an arrest, or a hospital visit, something urgent that needs money right now. The voice sounds right. The emotion sounds right. The caller ID might even show a familiar name.

By the time the family figures out what happened, the money is gone.

Why Connecticut? Why Now?

The CT News Junkie report doesn't point to a single reason Connecticut ranked first, but a few factors likely contributed.

Connecticut has one of the highest median household incomes in the U.S., around $90,000 according to Census Bureau data. Higher-income households have more assets, which makes them more attractive targets. The state also has a significant retiree population along the shoreline and in Fairfield County, and older adults are disproportionately targeted by impersonation scams.

Beyond demographics, 2025 was the year AI voice tools became genuinely cheap and accessible. What once required a studio and a skilled audio engineer now runs on a laptop, and the barrier to pulling off a convincing voice clone dropped to almost nothing.

The Bank Spoof Problem

Voice cloning gets the most attention, but spoofed caller ID is just as damaging. Scammers can display any number they choose when placing a call: your bank's fraud department, the IRS, a local police department. The number looks real because they made it look real.

The call follows a familiar pattern. There's suspicious activity on your account, they need to verify your identity, and they need you to move funds to a "secure" account to protect them. The urgency is manufactured, the number is fake, and the "secure" account belongs to the scammer.

The Federal Trade Commission reported that impersonation scams (government and business combined) cost Americans $2.95 billion in 2023. With AI making these calls more convincing, 2025 numbers are expected to be higher when full federal data is released.

What Makes This Hard to Spot

These calls are designed to bypass your skepticism. The voice is right, the number is right, and the story creates urgency that shuts down careful thinking.

Research on human detection of AI-generated audio is not encouraging. Studies have found that people identify synthetic voices correctly only about 55-60% of the time under controlled conditions. In a real call, with stress and emotion in the mix, that number almost certainly drops.

You can't trust your ears anymore. That's not a flaw in you. It's a flaw in a system that was built before AI voice cloning existed.

Three Words Change Everything

A private codeword challenge stops this kind of call cold.

Trust Onion gives you three simple words that rotate every 60 seconds. When a call comes in from someone claiming to be your son, your bank, or anyone else you're supposed to trust, you ask one question: "What are the words?" If they know them, it's really them. If they hesitate, make excuses, or give the wrong answer, you hang up.

The words are calculated locally on your device, with no server or internet connection required. An AI can clone a voice, but it cannot know three codewords that live only on authorized devices and change constantly. A spoofed caller ID tells you nothing about whether the person on the line is who they claim to be. The codewords do.

A Proofie™ is a selfie signed by your passkey the moment you take it. Share the link, and anyone can check in their own browser that your key signed this exact photo, after a known moment, and that nothing has changed since. A standard P-256 WebAuthn passkey signs after Face ID, fingerprint or device-passcode unlock. The signature must carry both user-presence and user-verification flags. This is not a crypto-wallet signature. Identity is the full signing public key, not its shortened label. Save a previously trusted key to recognize later Proofies from the same key. This does not automatically verify a real-world identity or detect AI-generated images. Signed time and place cannot be changed undetected; they are not independent proof of physical whereabouts or exact capture time. Family codewords work alongside Proofies, not as an overlay or signed field. The photo is permanently published to public IPFS, and the full signed record is public JSON pinned to IPFS. Removing a Proofie in the app hides it from your list and disables its share link, but does not erase the photo from IPFS. Only create Proofies whose photo and signed information you are comfortable making public permanently.

Trust Onion is free. Setup takes minutes. And the next time a panicked voice calls saying they need money right now, you have a simple, private way to know if it's real.

What People in Connecticut (and Everywhere Else) Can Do Right Now

Talk about this with the people you trust before it happens. Scams work partly because no one has a plan when the call comes in. The voice sounds real, the number looks right, and there's no agreed-on way to check.

Set up a shared codeword challenge with anyone you might need to verify: relatives, close friends, or anyone who might call you in an emergency. Trust Onion keeps the words rotating automatically, so there's nothing to memorize and nothing that can be stolen and reused.

Remind the people you care about: no bank, government agency, or legitimate organization will ever ask you to move money to protect it. That script, every time, is a scam.

If a call comes in that feels wrong, even a little, hang up and call back on a number you already have. Real emergencies can wait 30 seconds for you to verify who you're talking to. Fake ones can't.

Connecticut lost $16.1 million last year. That's 16.1 million reasons to have the conversation now.

A controlled voice-cloning safety demonstration

Try our tool to see just how easy voice cloning is. This controlled safety demonstration exists only to show how easy voice cloning is and to make the risk tangible. It is not a real incident or proof of anyone's identity. Your submitted sample is handled transiently to generate the fixed demonstration audio. Trust Onion does not identify you or the person from the sample. Trust Onion does not store the submitted recording, clone, or generated audio.

A lasting record of a place and time. Proofie™ images are stored on IPFS, a third-party distributed file network. Once created and shared, copies can’t be deleted everywhere by you, the recipient, us, or bad actors. So when you create a Proofie, you make a verifiable record of the place and time it captures.