Playground for Predators. Profit for Platforms. Who’s Really Grooming Our Kids?
Updated
In 2019, the online platform and game creation system Roblox told the U.S. government about 675 suspected cases of child sexual exploitation on its platform. Four years later, that number had grown to more than thirteen thousand. In 2025, that number had grown to over 65,000. Nobody had to go undercover and coax those figures out of a whistleblower. No indeed. Instead, the company reported them itself, year after year, while also telling parents that Roblox was a safe place for them to leave their kids.
That is the whole story in a nutshell. Everything else is detail. But our children are being exploited, and those details matter.
It is critical that we fully understand exactly who is on Roblox and similar platforms. It is largely children, after all. By Roblox’s own accounting, in the United States, two out of every three children between the ages of nine and twelve play on it. Some users are less than ten years old. Nebraska’s attorney general looked at Roblox and aptly called it a playground for predators. Texas filed suit in November 2025 under a headline about pixel pedophiles, which hints at how much patience is left in the offices of those who see the enormous problem here.
Nearly every complaint about the exploitation of children on these platforms has the same pattern. Initially, contact between a child and a predator occurs within the online game itself, where a child believes they are talking to another child. Once a conversation is established, it moves to another platform, usually Discord or Snapchat, where messages vanish and parents are left in the dark, not knowing how to follow along. Those destination platforms are seeing the same climb in crimes against our children. At Discord, its own transparency report showed child-safety violations rising by 150 percent in the last quarter of 2023 alone.
But it gets worse. If the past few years weren’t bad enough, there is now a rapidly evolving partner in crime targeting our children that has changed everything. This new aspect must be clearly spelled out, because coverage typically touches on it and then moves on. The new aspect at play is artificial intelligence, which has done three separate things to facilitate the sexual exploitation of children online, and each one is worse than the last.
The first is images. It used to require real skill with photo-editing software in order to make a convincing fake nude of a real person. Thus, it remained infrequent. But now, thanks to AI, that barrier is gone. Apps have been built for one purpose: to take an ordinary clothed photograph (in this case, of a child) and generate a naked version of the person in it. These apps are called nudification tools, and they work in a few clicks. And the source photo does not have to come from anywhere sinister. A school website or a parent’s public post will work just fine. In fact, more than thirty thousand reports last year involved a predator feeding a real child’s picture into one of these tools to manufacture explicit material. Once complete, the finished image is then sent to the child, along with a threat about who else might see it.
No physical contact ever had to occur. And to make matters worse, the child did nothing, sent nothing, and is now being extorted with a photograph of himself or herself that does not actually exist. And most of the time, the child will tell no one. Not because they are confused about whether the image is real, but because they know that nobody looking at it will be able to tell. Research by Thorn found that sixteen percent of these victims never disclose what is happening to them to anyone. According to the FBI, the reason for the silence is that children are afraid of what the criminal threatened, afraid they will be blamed, afraid their phone will be taken away, and are often told by the predator that they have committed a crime themselves and will be prosecuted for it. Sadly, ninety percent of the children caught in the financially motivated version of this nightmare are boys between the ages of fourteen and seventeen, and seventeen percent of victims were twelve or younger the first time it happened. At least twenty of these cases have been connected to a child’s suicide by the FBI.
The second damage inflicted by AI is conversation. The National Center for Missing and Exploited Children (NCMEC) has found AI tools being used to generate grooming dialogue, and its own assessment is that the technology can simulate the experience of an explicit chat with a child. Around three thousand reports last year involved AI used this way in chatrooms on gaming platforms. Think about it. What used to take a predator weeks of patient effort, grooming one child at a time, can now be run at scale by software that never gets tired and never gets bored. Thus far, no documented system has been found that selects, grooms, and delivers children to a waiting adult. What is documented is a human being using these tools to do by Tuesday what used to take until Christmas.
The third facet introduced by AI is volume, and it squarely impacts the people trying to help solve these crimes. Investigators hunting for children in real danger now have to sift real victims out of an ocean of synthetic images. And the platforms aren’t making it easier. Upon reviewing the material submitted to its tipline between 2023 and this year, NCMEC found more than ten times as many AI-generated abuse files as the companies had flagged as such. Under federal law, these platforms must report child sexual abuse material once they are aware of it, but nothing requires them to go looking for it in the first place, which is worth holding alongside the Roblox numbers mentioned in the opening paragraph, because thirteen thousand is not a count of what happened on the platform. It is a count of what the platform noticed. And when they do report, they frequently fail to say whether what they found was generated by software or if it documents the abuse of a real child, which is the single most important thing an investigator needs to know.
Without question, the trajectory in this nightmare is not gradual. Reports to the NCMEC tipline involving generative AI went from 6,835 in the first half of 2024 to 440,419 in the first half of 2025. In other words, sixty-four times higher in twelve months. Reports of online enticement, meaning an adult communicating with a child toward sexual ends, rose 77 percent over the same timeframe. By the close of last year, AI-related reports to the tipline exceeded 1.5 million.
These online platforms were not caught off guard by any of this. When Utah’s lawsuit against Snap was partially unredacted, it revealed that senior engineering managers had described the rollout of the company’s AI chatbot as reckless, with employees warning that it had not been adequately tested and could be prompted to say nearly anything. They shipped it out anyway. The same filing shows Snap internally acknowledging that its platform was overrun with sexual extortion.
Highlighting the debacle targeting our children, the reporting itself is also a problem. Eight companies, including Snap, Roblox, Discord, and Meta, filed over 17 million exploitation reports in 2025, accounting for 81 percent of everything the national tipline received. Sure, that sounds like diligence until you read what Senator Chuck Grassley’s oversight on the matter turned up. He found that the reports these companies file are frequently too thin for police to act on. A tip that arrives without the information needed to identify a suspect or locate a child is a number in a spreadsheet, not details that will help rescue a child.
Presently, the bill for this crime against children is now coming due in court. For example, roughly 170 cases have been consolidated in federal court in California, with more in state court in Los Angeles, and attorneys general are suing or investigating in a dozen states. Nevada has already settled for $10 million.
And, as expected, Roblox has an instant fix to keep kids in the game. Anyone who wants to talk to another person on the platform now has to prove their age first. The default method is to submit a selfie video so an outside company called Persona can scan their face and sort them by age. Government ID works too, for those who’d rather be tracked that way. In other words, either offer your child’s face or their papers, so they can say hello to a stranger in a game. Sigh. Roblox has had this type of age verification technology available since 2023. But what it lacked until the lawsuits arrived was any urgency to require it to protect children. When it finally did, the rollout reportedly cost Roblox tens of millions of users and billions in market value, which indicates precisely why it waited. And, by the way, we need to be convinced that these invasive measures will actually protect our kids.
Again, the industry that put generative tools in every hand on earth is now selling us the machine that will protect our children from the last machine. There is money in both (hey parents—an easy solution is to create a reality for your children that is offline). Yoti, one of Persona’s competitors, says it has run more than a billion age checks and has become profitable as governments have begun demanding them. The failure created a market, and no doubt helped facilitate online surveillance (which is another story in and of itself). These platforms will call it innovation, and the counting will go on. But will these extra measures actually protect our children?