Facebook Flooded with AI‑Made Child Abuse Videos; Meta’s Weak Response

Content warning: This article describes graphic, AI‑generated depictions of child abuse.

Over the past year, Facebook and Instagram feeds have become a grotesque gallery of synthetic cruelty. The videos aren’t filmed with a camera – they’re rendered by advanced AI models, yet they look unsettlingly real. Babies are slammed, toddlers are burned, and children are chained or forced to eat trash. The clips circulate in dozens of accounts that seem dedicated solely to posting this content.

One clip shows a little girl sobbing while a woman shouts, “Please don’t hit Noah, hit me instead.” The girl’s face is streaked with bruises, and another child cowers in the background. In a different video, a dirty‑looking boy begs for money for food, only to be slapped twice, blood running from his nose. Another scene has a boy tied up while hot wax drips onto his hands, his scream echoing the horror.

What makes these videos especially chilling is the pattern behind them. Many share the same grim motifs: a child denied an inhaler, a kid forced to eat shellfish despite an allergy, or adults dumping snakes and cockroaches on a crying youngster. Some portray Black children being beaten by white adults in a setting that looks like a colonial plantation. The violence is repetitive, suggesting creators are feeding a generator with specific prompts to churn out more of the same.

Even though the footage isn’t perfectly photorealistic, the quality is high enough to provoke strong reactions. Thousands of users comment, some horrified, some oddly fascinated, and a few even ask for “the next part.” The engagement numbers—likes, shares, and comments—sometimes climb into the tens of thousands, giving the posts a disturbing visibility boost.

We tried to map the network of accounts posting these clips, but the web kept expanding. New pages appeared faster than we could catalog them, indicating a near‑endless supply line. Meta’s own policy on violence against children is vague: it bans “real or non‑real non‑sexual child abuse” unless the material is from art, cartoons, movies, or games. The policy doesn’t explicitly address AI‑generated content, leaving a gray area that bad actors exploit.

When we reached out, Meta issued a boilerplate response: “We have strict rules against content that depicts physical harm to children—whether it’s real or AI‑generated—and we have removed these videos.” Yet the company’s public data on takedowns is opaque, and after flagging dozens of clips, many remained online—a girl locked in a chest freezer, a boy discarded in a trash can gnawing on bones, a child forced to eat garbage in a classroom.

Critics argue that Meta’s enforcement is more about optics than substance. Recent reporting also showed the company accepting money from advertisers to promote apps that generate AI‑crafted sexual imagery of minors, some of which were built from stolen photos. The pattern suggests a broader tolerance for AI‑driven exploitation, as long as it can be packaged as “advertising revenue.”

In short, the flood of AI‑generated child‑abuse videos on Facebook is real, the content is grotesquely detailed, and Meta’s reaction—while publicly condemning the material—has been slow, inconsistent, and largely ineffective.