Inside the DeTrade Fund Scam and the Deepfake Panic It Sparked

Fabricated news isn't a new problem, but the tools producing it have changed shape. Technology designed purely to deceive doesn't leave much room for a moral use case, and deepfakes may be the most unsettling version yet — a form of media manipulation that undermines confidence in our own eyes and ears. Once people have reason to distrust their own senses, verification stops being straightforward, and a fractured audience ends up picking sides on what's real. The information age has turned disinformation — or just the fear of it — into a constant background task of separating fact from fiction and marketing.

A scam built on a borrowed face

A recent incident put this exact problem of digital identity and trust on display. DeTrade Fund marketed itself as a community-governed project offering holders of its native DTF token access to centralized-exchange arbitrage bots.

The team ran a private token sale that pulled in 1,438 ETH — and then disappeared with it.

To earn early trust, the operation invested in constructing a convincing identity, both online and off. It registered an actual company with Companies House in the UK, and secured coverage through press releases on multiple outlets, each one repeating the face and name of its supposed chief executive, "Mark Jensen." In a market dominated by anonymous teams, having a "real" face out front set DeTrade apart — and exploited the trust people instinctively place in a human face.

Once the private sale turned out to be built on nothing, scrutiny shifted to the promotional videos DeTrade had circulated on social media. Suspicion grew after one user flagged that the man appearing as CEO looked like an AI-generated construct rather than a real person — a claim laid out in a Twitter thread that spread quickly.

An expert weighs in

Rekt sought out an anonymous specialist in synthetic media — referred to here as IE — to make sense of the claim.

IE described running a company that builds proprietary machine-learning systems for producing and distributing adult content, work that centers on deepfake methods for inserting AI-generated likenesses into video in a photorealistic way. Before moving into crypto-adjacent commentary, the team's decade-plus background in machine learning and computer vision had mostly served large-scale agriculture and pharmaceutical applications.

Asked for a first impression of the DeTrade footage, IE noted plenty of speculation calling it "AI generated," a deepfake, or both — but after multiple viewings, offered a more precise read: it may well be a deepfake laid over a real person's face, but the figure in the video is almost certainly not an AI-generated person. IE stressed those are two distinct technologies. Their own firm trains generative adversarial networks (GANs) to invent entirely new people who don't exist — something categorically different from a deepfake.

A deepfake, IE explained, still relies on AI, but not in the sense of manufacturing a new identity. It takes existing footage of a real face and overlays it onto someone else in an existing video, producing — when done well — a convincing faceswap showing someone doing something they never actually did. In other words, no new person is created; an existing face is simply grafted elsewhere. An "AI-generated person," by contrast, comes from training a GAN on large volumes of photos of real people until the model learns the general concept of a face or body, at which point it can synthesize new individuals that never existed — the reverse operation from a deepfake, incapable of inserting anyone into footage but able to conjure new people from scratch. Applied to DeTrade, the claims combined both ideas — AI-generated and possibly also deepfaked over existing footage — and IE reiterated that the AI-generation part is almost certainly wrong.

The reasoning: only around five organizations worldwide currently have the combined capability to generate a photorealistic person from nothing and then composite that person into existing video at this level of fidelity — and IE's company is one of them. The rest are either massive players like Nvidia or well-capitalized AI startups with tens of millions in funding. It strains credulity, IE argued, that senior staff at a company like Nvidia would risk their reputation to skim a few million dollars from the DeFi community. By contrast, plenty of people could have lifted the face in that video from ordinary existing footage — a professor's Zoom lecture, for instance — making a simple faceswap far more plausible than a from-scratch AI persona.

Asked whether tools like these mainly breed paranoia, IE didn't push back on the premise: some degree of paranoia is warranted. The core lesson, in IE's view, is that seeing a photo, video, or audio clip of someone is no longer proof that the person actually said or did what's shown, or even that they exist. Many people grasp this in theory but still treat a "non-anon" team — someone doing a livestream or posting a photo with an ID — as meaningful extra assurance, when it isn't. Very few outfits can currently fabricate a person from nothing and drop them into video, and IE's is one of them, but the barrier keeps dropping, and amateurs or organized fraudsters will eventually be able to replicate it. Absent an in-person meeting, IE argued, there's no reliable way to confirm someone is genuine or that footage of them is authentic — summed up as: without a handshake you can't be sure someone exists, and without holding their ID yourself you can't be sure it's real either.

A closing thought

Media manipulation technology has always drawn resistance, echoing older fears around new inventions, and opinions may well soften with time — though given the current coverage, that shift is hard to picture today. Salvador Dalí once wrote, "Si muero, no muero por todo" — "If I die, I won't completely die." Three decades after his death, that line has taken on an unexpected resonance at The Dalí Museum, where the sentiment is now illustrated in ways he couldn't have anticipated.