Billionaire investor Mark Cuban has publicly criticized Meta (formerly Facebook) for knowingly allowing AI-generated deepfake advertisements to target users. Cuban reported receiving numerous emails from individuals claiming they were scammed into purchasing a fake diabetes cure for $294. These advertisements allegedly use AI-generated copies of Cuban's face and voice to sell products he does not offer. Cuban noted that some of these ads even include disclaimers stating the content is AI-generated for entertainment, arguing that the scam succeeds before any viewer reaches the disclaimer. He characterized the platform's tolerance of such content as a defining cultural issue for the organization.
The scale of this issue is highlighted by internal Meta documents reported by Reuters. These documents indicate that Meta expected roughly $16 billion—approximately 10% of its 2024 revenue—to flow from advertisements promoting scams and prohibited products. Meta's internal safety staff estimated that users across its apps saw roughly 15 billion "higher risk" ads per day, which are ads with obvious markers of fraud. While Meta's automated systems are designed to identify these ads, the documents reveal a specific moderation threshold: before a platform bans an advertiser, the automated system must peg the account as at least 95% likely to be running a scam. If a suspected scammer falls below that 95% threshold, the system allows the account to stay live, provided the advertiser pays a higher fee per advertisement.
This moderation policy has significant consequences for vulnerable populations. According to the FBI's 2025 Internet Crime Report, Americans aged 60 and older lost $7.7 billion to internet crime in 2025, a 59% increase over the previous year. Of these crimes, investment schemes were the most damaging, while tech and customer support scams accounted for over $1 billion in losses. Notably, AI was involved in more than 3,100 of these complaints, with victims losing over $352 million in AI-related fraud. In response to the scrutiny, a Meta spokesperson stated that the $16 billion figure is "rough and overly-inclusive" and noted that the company removed more than 134 million scam ads in 2025. However, the Securities and Exchange Commission is currently investigating Meta regarding advertising tied to financial fraud.
One significant uncertainty remains regarding the exact amount of revenue generated by "knowingly" permitted fraud versus ads that bypass filters due to technical limitations. Because the $16 billion figure is an internal estimate described by Meta as "overly-inclusive," it is difficult to determine the precise proportion of revenue derived from ads that Meta's systems intentionally allowed to run under the 95% threshold rule versus those that were missed entirely by the automated detection systems. This distinction is central to understanding the trade-off between aggressive fraud prevention and the maintenance of high-volume advertising revenue.