Stolen Voices: The AI Fraud Wave Knocking on Vietnamese Football's Door

Stolen Voices: The AI Fraud Wave Knocking on Vietnamese Football's Door On July 26, Samuel Sarmiento — former lead vocalist of the Mexican banda group Banda Los Recoditos — announced on his personal page that his mother had just passed away. Within hours, an audio recording of a voice identical to his began circulating in fan chat groups of regional Mexicano music. That voice urged people to send money for the funeral. Sarmiento was forced to publicly deny it: he had not called for donations, and anyone receiving such a request was being defrauded. I read that news three times, not for its content, but for its structure. A famous person had just lived through a real event. The fraudsters latched onto that real event within hours. They did not need to invent a story; they only needed to borrow a fact that was hot, then attach the victim's voice on top of it. That was when I understood this problem no longer belonged to the Mexican music scene. It was knocking on football's door. The xG shock at Hang Day turned me from a spectator into a reader of data. But there is a kind of data that the xG table can never measure: data about trust. This time, the attacker is aiming straight at the biggest noise variable in all my models — trust. CONTEXT: ELEVEN SECONDS OF AUDIO IS ENOUGH In 2026, to build a convincingly cloned voice, an engineer needed hours of high-quality recordings, several thousand dollars, and weeks of model training. By 2026, that threshold had fallen to a few minutes of audio. By 2026, some commercial tools needed only about eleven seconds of a sample. The cost of a single voice synthesis is now lower than a bowl of pho in Saigon. Technology always becomes cheap before ethics does. That is a rule I learned from years of watching the betting market operate. When a tool becomes cheap and accessible, it gets used for bad purposes before good ones, simply because bad purposes have clearer financial motives and shorter payback cycles. What made me pause at the Sarmiento story was not the technology. The technology had been there for a long time. What made me pause was the fraudster's operating principle: they do not create the event, they only parasitize a real one. A singer who just lost his mother is a real event, with a date, with confirmation from the person himself, with tributes from colleagues. The fraudster does not need to build context; they only need to insert a voice into context that has already been authenticated. And here is the part that made me think of Vietnamese football. Football is the environment that produces real events at the highest frequency of any entertainment industry. Injuries are real. Transfers are real. Suspensions are real. A player losing a family member is real. A club facing financial trouble is real. Each such event is a piece of context a fraudster can latch onto, and each famous player is a voice they can clone in eleven seconds. I do not predict the future; I only read ahead into the way the past keeps operating. And the past of the fraud industry shows one clear rule: wherever money flows through emotion, the fraudster arrives first. WHY FOOTBALL IS A NEAR-PERFECT TARGET Let me build this problem as a model, the way I always do with a match. Variable one: audience scale. Vietnamese football has one of the most passionate fan bases in Southeast Asia. A national team match can draw millions of television viewers and hundreds of thousands of social media comments within ninety minutes. To a fraudster, these are not spectators — they are a pre-sorted list of potential targets ranked by emotional intensity. Variable two: attention concentrated on individuals. Fans do not just follow the team; they follow every player, every player's family, every smallest move. A player posts a wedding photo, a player announces a child's birth, a player shares a sad story — all become breaking news within hours. The fraudster does not need to gather data; the fans and the media have already made it public for them. Variable three: a low threshold of skepticism. This is the variable I care about most, and also the most dangerous. In football, fans are trained to believe in the unexpected. They believe a weak team can beat a strong one. They believe a player can shine in a single moment. Belief in miracles is the essence of this sport. The fraudster exploits exactly that essence. Variable four: easy payment systems. Bank transfers, e-wallets, QR codes — all allow money to move within seconds. There is no waiting period long enough for the victim to sober up. Variable five, and this is the least discussed one: football has a murky financial ecosystem. Betting, transfers, sponsorship, youth academies — all are areas where money does not always move transparently. A fraudster impersonating a player who needs capital for an investment, or impersonating an agent who needs a deposit for a transfer deal, blends into an underground economy that already exists. Based on my experience watching matches, I realized one thing about football that few people notice: its strength lies in collective emotion, but its weakness lies there too. A tightly bonded community is a community vulnerable in a contagious way. NOW FOR THE DATA I always tell readers that a sure thing does not exist; there is only probability mispriced and sold correctly. AI fraudsters operate on the same pricing logic. They do not target randomly; they target what is mispriced in terms of trust. Imagine a player with two million followers. If a fraudster sends fake messages to a small fraction of them, and only a tiny fraction believe it, the absolute number is still large enough to profit. This is a simple expectation calculation: expected profit equals probability of success times the amount collected, minus cost. With a cost near zero, a success probability just above one percent is enough to be profitable. That is why I do not call this a scam. I call it an industry with a stable business model. And when a business model is stable, it expands into new markets. That new market, with all the features I just listed, is football. Let me redraw the operating model of an AI voice fraud targeting a player, based on the structure I observed in the Sarmiento case. Step one: choose a target with a hot event. Fraudsters monitor news in real time. A player just suffered a serious injury, a player was just fined, a player just had family news — that is the golden moment. Step two: create a fake voice. With eleven seconds of public audio from a television interview, the fraudster creates a copy convincing enough to deceive a listener over the phone or in a voice message. Step three: choose the approach channel. Fan chat groups, private messages to the most devoted fans, player support groups on social media — places where the threshold of skepticism is lowest. Step four: create an urgent story. A sum needed quickly for a moving purpose — funeral costs, helping family, a charity project the player had not yet announced. Step five: collect the money and disappear. Once the money is in, the account is deleted, and the real player must publicly deny it — but the money is already gone. I realized this structure is identical to a betting model with a house edge. The fraudster does not need to be right every time; they only need to be right often enough to profit at scale. THE CONTRARIAN ANGLE: TECHNOLOGY IS NOT THE CULPRIT This is the part I want to spend the most time on, because I believe most writing on this subject is focusing on the wrong thing. When an AI scam occurs, the media's first reaction is to blame technology. Headlines about "deepfakes," about "dangerous artificial intelligence," about "the post-truth era" appear densely. But if we read the Sarmiento case closely, we see something different: technology is only the vehicle, while the real vulnerability lies elsewhere. That vulnerability is the verification gap. Think about how we verify information in football. When a transfer rumor appears, we check the source. When a player posts an announcement, we check the official account. But when a voice speaks over the phone, we have no verification process at all. Voice has always been treated as evidence of real presence. We believe voices instinctively, without mediation. That is the biggest blind spot of this era. We have built verification systems for images, for text, for account identities — but not for audio. And here is my second contrarian angle: it is the fans' attention itself that is the exploited variable, not the technology. Without passionate fans, there is no one to defraud. Without curiosity about players' private lives, there is no context for the fraudster to latch onto. Every time we share a player's personal information without verifying it, we are unintentionally widening the attack surface. I call this the reputational attack surface. The more famous a person is, the larger their surface. And in football, where fame is built from collective emotion, the attack surface is larger than in any other industry. What is interesting is that Sarmiento's response was strategically sound. He did not just deny; he proactively warned fans before they became victims. He understood that in a voice attack, time is the decisive variable. The faster the denial, the fewer people are defrauded. This is a lesson Vietnamese football clubs should remember. Trust is a noise variable; run an emotion regression before you place a bet. I say this to readers about betting, but it applies perfectly to this subject. Before believing a voice, run a simple regression: does this voice match that person's usual behavior? Does this person habitually solicit money through personal channels? Is there a second source confirming it? Three questions, three verification steps. No complex technology needed. Only discipline. AND HERE IS THE HARDEST PART: WE ARE MEASURING THE WRONG THING There is a data problem I must state plainly, because it concerns how we understand this fraud wave. Every fraud statistic is based on reported cases. But most fraud cases are never reported. Victims are ashamed. Victims do not know they were defrauded. Victims are famous people who do not want their names tied to a scandal. This means the true number is always larger than the number we see, and we do not know by how much. In statistics, this is survivorship bias. We only see the cases that surface. The ones sunk below do not generate data. And when we build policy on biased data, we react slowly and in the wrong direction. For Vietnamese football, this means we should not wait for a major fraud case to happen before acting. We should build prevention now, based on the model rather than the incident. I remember the Kazan shock. Kazan does not take revenge; Kazan only keeps the books and waits for me to miscalculate. I had published a prediction that Germany would be eliminated in the 2026 World Cup group stage, based on declining pressing data, and was mocked. But data does not care about mockery. It only operates by its own rules. On the night of June 27 in Kazan, Germany lost to South Korea, and my model held. The lesson here is: sometimes data points to a risk before the risk occurs, and being mocked does not make the risk disappear. The AI fraud wave is a risk that data is pointing to before it crashes into Vietnamese football. WHAT WILL HAPPEN IN THE NEXT CYCLE I do not predict the future; I only read ahead into the way the past keeps operating. And the past of fraud waves shows a pattern of geographic expansion. It begins where there is technology and money, then spreads to where there is emotion and fans. In the next cycle, I think probability leans toward three scenarios. First, fake player accounts will become more sophisticated. Not just fake social media accounts, but voice messages, fake calls, short videos with synthesized voices. Fans will find it harder to tell the difference. Second, clubs will have to build official verification processes. A single contact channel, a confirmation code, a public process for fans to check any financial request made in a player's name. Third, and this is the scenario I worry about most, fraudsters will target players' families, the people least trained in media and most vulnerable. Every analysis of mine now has a section behind the numbers, because I have realized that behind every metric is a human being. Behind every AI voice fraud is a fan who placed complete trust in someone they loved, and was betrayed by that very trust. That is what I want to leave. Not a conclusion, but a question for the next cycle: when technology can clone anyone's voice in eleven seconds, what is left for us to authenticate a human being? The answer, perhaps, is not in technology. It is in discipline. In the habit of checking the source before believing. In building contact channels that fraudsters cannot fake. And in admitting that trust — football's most precious asset — is becoming the most valuable target of those who hunt money. At 59, I have a perspective: every cycle is a loop with a remainder. This fraud wave will pass, but its remainder — the healthy skepticism it leaves behind — may be what football needs to mature. The day the model breaks is the day the data monk must burn it down and start again from the original scripture. And the original scripture this time has a new chapter: the chapter on protecting trust in an age when a voice is no longer proof of identity.