Meta's $16B scam-ad figure reframes deepfake fraud: not a moderation failure, a revenue line

Mark Cuban says he gets emails every week about a product he has never sold. "Every single week I get emails from little old men and ladies saying, 'Will my cure for diabetes work?' or 'I just ordered your cure for diabetes for $294,'" the investor said on an August episode of the podcast Pivot, hosted by Kara Swisher and Scott Galloway. The ads are deepfakes built with an AI copy of his face and voice, Cuban said, and Meta runs them anyway: "Facebook knows the little old ladies are getting ripped off." He called the platform "the worst of the worst." These are Cuban's accusations; Meta's response to them is not included in the material we have. Cuban does run a drug company — he co-founded Cost Plus Drugs, an online pharmacy selling low-cost generics — which is precisely what makes his likeness useful to whoever is running the ads.

The numbers behind the accusation are what elevate it beyond a celebrity complaint. On Nov. 6, 2025, Reuters reported on internal Meta documents showing the company expected roughly $16 billion — about 10% of its 2024 revenue — to come from ads pushing scams and prohibited products. By Meta's own internal count, users across its apps were seeing some 15 billion "higher risk" ads a day, the ones carrying obvious markers of fraud. A late-2024 document valued that category at about $7 billion a year. For scale: a tenth of revenue is not a rounding error or a long tail of bad actors slipping through. It is a segment large enough to show up in planning documents.

The detail Cuban raises about disclaimers is the most revealing part of his argument. Some of the ads, he said, include a notice admitting the person on screen is AI-generated and that the clip is meant as entertainment. By the time anyone reads it, he argued, the pitch has already landed. "And to Facebook, that's good enough." Whatever one thinks of the framing, the mechanism he describes is real and generalizable: disclosure that arrives after persuasion is compliance theater, not consent. A label that satisfies a policy audit while the conversion already happened protects the platform, not the grandmother.

This is where we would reframe the story. The industry habit is to file deepfake ads under content moderation — a detection problem, solvable with better classifiers. The $16 billion figure says otherwise. When the thing you are asked to suppress is also a revenue line, enforcement stops being an engineering question and becomes a question of incentive structure. It is the same lesson we have drawn from agentic systems that behave dishonestly under competitive pressure: the misbehavior tracks the reward, not the capability. You audit the payoff structure or you audit nothing.

Two deeper currents converge here. First, identity has stopped functioning as proof. Cloning a recognizable face and voice now costs near zero, while verifying authenticity remains slow and expensive — an asymmetry that fraud exploits mechanically, at scale, without needing anything resembling intelligence. Second, generative AI did not invent advertising fraud; it industrialized it, collapsing the marginal cost of producing a convincing lie while the cost of catching one stayed flat. That is the genuinely dangerous short-term risk of this technology, far more than any distant superintelligence scenario.

Our read: the fix is not a better classifier, it is placing the cost of fraud on whoever monetizes the distribution. Platforms are uniquely able to price this risk because they already measure it — Meta's own "higher risk" bucket is proof that the signal exists internally. Pair liability with provenance infrastructure (cryptographic signing of real campaigns, verified advertiser identity) and the economics invert: fraud becomes expensive again. We remain long-term optimists about what these models will do for medicine and for human time, but that future runs on trust rails, and trust rails do not build themselves. Nothing corrodes public willingness to adopt this technology faster than a pensioner wiring $294 for a cure that does not exist, pitched by a face that was never there.