By the time Lynn Phaneuf realized the man on his television wasn’t really his country’s prime minister, his money was gone.
Phaneuf is a Canadian retired teacher, the sort of careful investor scammers like. He and his wife mostly use their smart TV to watch the news. So, when what looked like a current affairs segment by the national public broadcaster appeared on screen one evening – complete with CBC branding and a familiar studio – he paid attention.
On screen was Mark Carney, Canada’s prime minister, apparently talking up a new government-linked opportunity for ordinary Canadians to invest in cryptocurrency. The clip aligned with what a group of “financial advisors” had been telling Phaneuf over the phone for weeks.
“It seems so real,” the septuagenarian said later, as he told his story to the real CBC.
That was the point. The Mark Carney on the screen wasn’t Carney. It was an AI-generated impersonation built from real footage of the politician, stitched into a fake interview, and hosted on a website that mimicked CBC’s layout and design but lived at a different domain. Regulators in western Canada had already flagged versions of the same video as part of a larger fraud campaign, but none of that was visible from Phaneuf’s sofa.
Over several weeks, the callers used Canadian phone numbers, walked him through a seemingly legitimate trading dashboard and even arranged for roughly C$800 (US$578) to land in his bank account – a staged “profit” designed to prove the system worked. Eventually, Phaneuf wired about C$2,800 (US$2,023) to a crypto platform he believed was tied to the scheme. When he tried to withdraw or close his account, the responses slowed and then stopped altogether. The site vanished and the advisors stopped picking up.
Securities regulators have issued a string of alerts about similar scams using AI-generated Carney videos and fake news pages to promote everything from “AI trading platforms” to offshore exchanges.
His story sits at the small end of a much larger trend. Crypto has always attracted scams – AI has simply handed scammers a better costume. What used to be a collection of clumsy internet jokes – distorted celebrity faces and glitchy videos – has turned into a production line of convincing forgeries featuring prime ministers, billionaires and tech chiefs, all apparently desperate to help you get rich.
In Canada, provincial regulators say reported losses to crypto scams in Saskatchewan alone reached more than C$1.3mn (US$940,000) by early November. Nationally, the Canadian Anti-Fraud Centre has logged hundreds of millions of dollars in reported crypto-related losses since the start of 2024, while warning that only a fraction of victims ever come forward.
“If you’re opening up your search engine and you’re searching up a crypto investment, the first five or ten are going to be more than likely fraudulent platforms,” a spokesperson for the agency said earlier this year.
Globally, from 2020 to 2024, estimated scam activity grew by roughly 24 per cent per year, according to analytics firm Chainalysis. In its 2024 crime report, the firm said crypto scams had taken in at least US$9.9bn onchain that year and warned annual totals could exceed US$12bn.
“GenAI is amplifying scams, the leading threat to financial institutions, by enabling high-fidelity, low-cost, and highly scalable fraud that exploits human vulnerabilities,” said Elad Fouks, head of fraud products at Chainalysis. Generative AI “facilitates the creation of synthetic and fake identities, allowing fraudsters to impersonate real users and bypass identity verification controls,” Fouks said.
Fake news bulletins and bogus keynotes
Law enforcement is starting to catch up and is increasingly finding deepfakes at the front end of some of the largest crypto frauds on record.
On 4 Dec, Europol announced it had helped dismantle a European crypto scam network accused of laundering an estimated €700mn ($816mn). According to the agency, the group used glossy online investment portals and boiler rooms – and, crucially, AI-generated videos “impersonating renowned media outlets, celebrities and politicians” – to funnel victims into these fake platforms.
In October, viewers searching YouTube for Nvidia’s GTC developer conference discovered a channel calling itself “Nvidia Live”, streaming what looked very much like the real thing. On stage was Jensen Huang, the company’s chief executive, talking through a crypto “mass adoption event.” QR codes for a supposed token distribution flashed on screen. The only problem was that Huang wasn’t actually saying any of it.
The genuine keynote, on Nvidia’s official channel, drew a standard tech-conference audience. The fake one, featuring a phoney CEO, reached nearly 100,000 viewers before it was taken down.
Elon Musk has been the face of similar tricks for years. Early attempts laid fabricated audio over preexisting footage, having the billionaire promote “special Bitcoin giveaways.” The technology has since matured, becoming far more convincing and easier to create, while the underlying bogus platforms and promises have not.
In January, a campaign branded around an “ELONBTC17” promo code used AI-generated videos of Musk to promote a Bitcoin giveaway on YouTube and social platforms. The videos looked legitimate at first glance. The trading sites they linked to looked polished. Victims were told to deposit crypto to unlock “rewards” that never existed.
Hijacked accounts and phantom tokens
Scammers increasingly sidestep the hard work of building their own audience. Instead, they break into existing, verified accounts and push deepfake content straight on someone else’s followers.
In February, hackers broke into the X account of Tanzanian billionaire Mohammed Dewji and used it to promote a brand-new token, “$TANZANIA”. They posted a short video of Dewji enthusiastically endorsing the coin. The man in the video smiled like Dewji, sounded like Dewji, and told his followers this was their chance to get rich. Investigators say the pump-and-dump scheme netted roughly $1.48mn before the scammers cashed out and wiped most of the posts. A fabricated image lingers today of Dewji grinning, holding up a seemingly handwritten note reading “$tanzania”.
By mid-year, Brad Garlinghouse, chief executive of Ripple – the issuer of payment token XRP – had his own digital dummy. An AI-generated video of Garlinghouse began circulating on YouTube and X, announcing a “100 million XRP” rewards airdrop to celebrate Ripple’s legal wins. In the clip, Garlinghouse appeared to invite viewers to send XRP in order to “claim” their bonus. The accounts that shared the videos looked like Ripple’s own channels – some had been hijacked, others were cloned.
Deepfakes are moving up the food chain
For years, most of this effort was aimed at retail investors and the elderly – people like Phaneuf, sitting at home in front of a television, trusting the logos and faces on the screen. AI is now enabling fraudsters to climb the ladder.
In February, data firm Inca Digital exposed a fraudster who had been using “AI face changing tools” to sell bogus FTX bankruptcy claims to trading firms. Posing as an FTX customer under multiple aliases, he approached two companies that buy claims, offering to sell them sizeable positions. FTX debt had quietly turned into a niche asset class – risky but potentially lucrative – attracting distressed-debt desks, specialist funds, and, inevitably, someone willing to sell claims they didn’t actually own.
Inca’s report shows the perpetrator stitched together an artificial persona using newly created email addresses, forged Singaporean ID cards, and an AI-altered face on video calls that seems to be built on photos of jailed esports star Kurtis “Toyz” Lau Wai-kin. The deepfake-style video was good enough for due-diligence calls that convinced counterparties they were talking to a genuine creditor, even though Kroll’s KYC records later showed he was never an FTX customer at all. By the time the deception surfaced, more than $5.6mn in Circle stablecoins (USDC) had been routed through fresh wallets and pushed into offshore exchanges.
At a blockchain conference in Dallas, Texas this year, defence attorney Carlo D’Angelo described another case, involving a friend of his – a well-known crypto commentator with newsletters, talk forums, and a large following.
“He had to post the embarrassing news that he got a call from an investigator,” D’Angelo said. One of the commentator’s clients had been scammed out of $4mn by someone pretending to be him.
“They went to the extent of deepfaking his identity. His driver’s licence, his wife’s identity, his children’s identity to prove to the suspect that was being duped that they were the actual person,” D’Angelo added. “They social engineered their way in. So, the biggest point of failure, I hate to tell ya, is you.”
“AI has the ability now to replicate voice, face, they can get on Zoom conversations with you, and you think you’re talking to a person who is not the person you’re talking to,” the attorney noted. “It’s sophisticated, so you’ve got to be nimble, and you’ve got to be very vigilant.”
The end of seeing as proof
Creating a deepfake is straightforward. Take a public figure who often speaks publicly – Musk on stage, Garlinghouse in interviews, Carney at press conferences – and harvest clips of their voice and face. Feed those samples into voice-cloning and face-swap software. Type whatever script you need and let the machine do the rest.
The deepfake itself is just one layer of the scam. It sits on top of hijacked or impersonated social-media accounts, scripted call-centre operations, cloned dashboards displaying fake returns, and a tangle of bogus “taxes” and “fees” that keep victims sending money long after their first deposit.
For people like Lynn Phaneuf, the cost is counted in thousands of dollars. For institutions, it can be millions. For everyone, the loss is truly symbolic. Formats that used to be bullet-proof evidence of the truth – a TV news segment, a Zoom call, a keynote speech – no longer suffice.
Once, 'I saw it with my own eyes' was the end of an argument, even on-screen. In the age of AI-powered scams, it’s hard to know what to believe anymore.
