Multiply all 144,041 reporters by the Federal Trade Commission's $10,560 median and you account for roughly $1.5bn, which is about 19% of the $8bn figure New York is citing. The other four fifths sits with a much smaller group, whose average works out near $55,500 a head once you divide the total by the reporter count, and that is not a household losing beer money to a slick advert; that is savings moved out over weeks. Which matters for where a dollar of prevention goes, because consumer education aimed at the median depositor addresses a fifth of the money.
The alert calls the $8bn AI-powered. The series underneath it counts investment-related losses of every kind, and nothing published puts a number on how many of those reports involved a cloned voice or a fabricated celebrity endorsement. The 38% is doing double duty in the same document, described both as a jump in the count of reporters and as a surge in fraud, and those are different series: 144,041 grown 38% implies about 104,400 reporters in 2024, while $8bn grown 38% implies about $5.8bn and an increment of roughly $2.2bn. Both readings are defensible from what was released. They are not the same claim, and a platform being asked to absorb liability will notice.
Or rather, the more interesting version of the distribution question: to move any burden onto the feeds, dating apps and inboxes that carried the creative, you need a per-surface denominator, and the only per-surface signal on offer is that dating-app victims lose more per head because the trust-building phase runs longer. This is probably wrong, but I read a consumer alert telling people to verify platforms against official regulatory registries as an admission of instrument scarcity. Secretary of State Walter T. Mosley's line, that anything looking too good to be true almost certainly is, is free to issue. Writing a rule for ad networks is not, and the Department of State did not do that.
Notice where the last dollar actually leaves. The counterfeit exchange, complete with live price tickers, extracts a final payment by inventing a tax or fee at the withdrawal screen, or it simply disappears, and that step is a payments question rather than a content question. The institutional version sits somewhere else again: the tech chief executive charged in 2025 with raising more than $40m by misrepresenting what the company's AI could do needed investors who never tested the product.
What would change my read is an FTC breakout showing most 2025 investment-loss reports involved synthetic audio or video, which would turn the deepfake attribution into a measurement. Until then, the defensible statement is that investment fraud got dearer, and AI is the most plausible explanation anyone has offered for the increment.
