Deepfake Crypto Scams
When Seeing Is No Longer Believing

A livestream turns up in your feed. On screen is a crypto personality you actually follow, the voice sounds right, and they're calling a trade with total conviction. Or your phone lights up and it's "support" — on camera this time, badge on a lanyard, a call-center wall behind them. Or the person you've been talking to for four months on a dating app finally agrees to a video call, and the face matches every photo you've seen.

Here's the problem: all three of those pictures can now be computed. Face swapping and voice cloning stopped being movie plot devices and became standard equipment in the fraud toolkit. The face on screen can be generated frame by frame, and the voice can be cloned from a few seconds of audio lifted off a podcast, a livestream, or a voicemail greeting. If you're new to crypto, "I saw him myself" and "I heard her say it" have quietly stopped being protection.

This piece covers three things: the shapes this fraud actually takes in crypto, what the documented cases look like, and the one line you should be defending — which is probably not the line you expect.

Three shapes this already takes in crypto

1. Fake celebrity and influencer livestreams shilling coins

Fake trade calls used to mean stolen profile photos, fake groups and doctored screenshots. The current version is nastier: the stream itself can be a synthesized face over a cloned voice, with a chat feed and a countdown staged to feel like something happening right now. The most industrialized form is the hijacked video channel — someone takes over an account with a large subscriber count, renames it after a well-known company, loops a deepfake of a famous founder, and overlays a promise to send back double whatever BTC or ETH you deposit to the address on screen. These have run for years, usually timed to real product launches and rocket launches, because that is when attention peaks. Michael Saylor, one of the most cloned faces in the industry, has said publicly and repeatedly that any video of him giving away bitcoin is a fake and that his team is constantly having them removed.

The damage isn't done by the technology; it's done by transferred trust. You would never click an anonymous link, but once you believe it's that person calling the trade, you stop interrogating the project and the promised return. So keep the two questions apart. Whether a public figure's likeness gets stolen is their problem; whether you act on a coin call from anyone at all is yours — and the answer to that one never changes. Following someone else's calls was already near the top of the list of first-year beginner traps; AI has only polished the shell.

2. Fake support that will now get on camera

We've written about telling real apps and support from fakes: spoofed download pages, impostor support handles, the "dedicated channel" you're asked to install. All of that is still running. The upgrade is that the impostor will now video call you. The face matches the badge, the room looks like a support floor, and they may offer to "remote in" and walk you through it.

Once video is on the table, most people quietly relax: would a scammer really show his face? He would, because it doesn't have to be his face. The objective hasn't changed at all — move your assets to a "safe account," submit your seed phrase so they can "verify" it, or sign in on a cloned app. However real the picture looks, real support at a major exchange will never ask you in a private chat to move funds, and will never ask for a seed phrase or a login code. Identity gets verified in exactly one place: inside the official app, or through the official site you typed in yourself. Handing over a seed phrase is handing over the keys — see how seed phrases actually get stolen.

3. Romance-investment scams that will finally turn the camera on

You may already know the pig-butchering script: months of relationship building, then investing, copy trading, or an "insider allocation" as the reason to send money. The standard advice used to be "insist on a video call." That advice now needs half a sentence added: they passed the video call, and you still don't send the money. Agreeing to video — even checking in on camera every day — only proves the persona is being maintained. It proves nothing about the platform. The moment "let me manage this for you," "insider allocation," or "a coin that can't lose" enters a relationship, treat it as a scam, rather than spending another week trying to establish who is on the other end of the call.

In the reported cases, this stopped being hypothetical

What follows is drawn from cases that were publicly reported and acknowledged, so you can check them yourself rather than take our word for it.

The reference case is the one at Arup, the British engineering firm. In early 2024, Hong Kong police disclosed that an employee in the company's finance function paid out roughly HK$200 million — about US$25 million — after joining what looked like a routine video conference. Every other participant, including the person presenting as the group's chief financial officer, was a digital re-creation. Arup later confirmed publicly that it had been targeted (CNN, 2024-05). Note what actually failed there: not the employee's intelligence, but a process that allowed a payment to be authorized inside a video call.

Voice-only versions are cheaper and just as effective. The FBI's Internet Crime Complaint Center has issued public warnings that criminals are using generative AI to make fraud more believable — cloned voices for emergency "it's me, I need money now" calls, AI-written profiles and messages for romance and investment fraud, and synthetic images to dress up nonexistent platforms (FBI IC3 alert I-120324-PSA, 2024-12). A few seconds of audio from a podcast, a livestream, or an outgoing voicemail message is enough raw material.

The pattern across all of them is a shrunken decision window. While you're still thinking "hang on, could this be synthetic?", the transfer has already left. And crypto is the preferred settlement rail for exactly the reason you'd guess: it moves across borders in minutes and it doesn't reverse.

What to actually do — defend the boundary, not the detection

Detection tests age badly. What catches a fake today may fail next year, so don't build your defense on "can I tell it's fake?" Build it on "what do I refuse to do, real or not?"

First: liveness checks on a video call — useful, not a safe

If you're on camera with someone, ask for a few things synthetic video still struggles with: a slow head turn to each side; a hand passed in front of the face and taken away again; several seconds of speaking in profile; an off-the-cuff answer to a question they couldn't have prepared, delivered with gestures. Many face-swap setups still smear, jitter, or fail to stay stuck to the face during occlusion, extreme angles, and fast movement. Lag, features that drift out of place, or a flickering edge around the jaw are enough — end the call. But treat this as raising the bar, not sealing the door. Even if they pass every test, all you've learned is that the call looks convincing. You've learned nothing about the coin they want you to buy.

Second: agree on a code word with your family and close friends

For anything involving a large transfer, an emergency, or a deadline that "expires tonight," confirm through a private question or phrase agreed in advance — and call back on a different channel you already trust, instead of settling it inside the suspicious call. An AI can clone a face and a voice; it has a much harder time knowing the running joke from your last family dinner.

Third, and this is the core of the whole article: the second anyone pitches a coin, it's over

You don't need to establish whether the person is real or synthetic. You don't need to check whether the influencer really appeared on that stream. You don't need to install the "official group" they sent you. The pitch itself is enough to walk away from. Handing your decisions to someone else's calls was a high-risk act long before deepfakes existed; AI has only made the wrapper more convincing. Holding the line "I take coin tips from nobody" is worth more than ten techniques for spotting a face swap. You don't have to beat the AI — you just have to hold that line.

Fourth: verify any support contact through official channels only

Install and update apps only from sources you already trust. If a "support agent" reaches you by private message, video, or an outside link, close it and start a new conversation from inside the official app's help center or the published channels on the official site. If they try to stop you leaving the current window to verify independently, that resistance is the answer. The finer details of spotting cloned apps and impostor agents are in the piece linked above.

The wider point is that keeping your activity on regulated, self-managed rails removes most of the situations where someone can lead you by the hand. If you don't have an account yet, open one yourself on Binance's own site (invite code BN1606) and do the setup and first steps under your own control, rather than letting a face on a screen tell you which button to press.

If the money is already gone — and one last thing

If a transfer has already left, cut the losses first: stop sending anything more, preserve the evidence quickly (chat logs, transaction hashes, the counterparty's accounts, any recording or screenshot of the call), then report it and work whatever paths the platform and the chain still allow. In the U.S. that means filing with the FBI at ic3.gov and the FTC at reportfraud.ftc.gov; elsewhere, your national cybercrime or fraud reporting line. Be extremely suspicious of anyone who surfaces afterward offering to recover your funds — that is a well-worn second scam aimed at people who have just lost money. We have a piece on whether scammed crypto can be recovered; working through it in order beats improvising.

So, back to where we started: the face in the livestream, the agent on the video call, the person who finally turned the camera on. Since face swapping and voice cloning became cheap, every one of those images has been watered down as evidence. Your job isn't to become a forensic analyst. It's to take the decision back: no coin tips from anyone, no relaxing your guard over funds and seed phrases just because there's a face on screen, and no large moves made under emotional pressure or a countdown. Not sure about a specific message or "opportunity"? Run it through our scam checker before you decide. Seeing is no longer believing — and simply accepting that puts you a step ahead of everyone who trusted the face.

Frequently asked questions