Meta leaves AI videos of violence against children on Facebook

A removal that did not stick

In the earlier case, journalists identified an account posting particularly graphic videos, including scenes of children being beaten and forced to eat garbage. Meta eventually took the account down. After the report was published, the account and its videos returned. A company representative said Meta was looking into the matter but offered no timeline; the account later disappeared again.

The latest test suggests that this was not an isolated moderation lapse. Journalists sent Meta 19 accounts yesterday. More than 24 hours later, each was still accessible. The company did not answer the follow-up request for comment.

The videos included depictions of:

The journalists did not publish the videos themselves. Other clips depicted Black children being beaten and humiliated by teachers. One showed a teacher striking a crying student with a ruler and asking why he did not attend a school for Black children.

Many videos end with an adult intervening to help the child. Meta had cited such endings in the earlier report as a reason the videos did not need to be removed.

The capacity is there

Meta could look for the detailed captions uploaded with the videos, which often describe the abuse. The caption for the oven video, for example, says a mother put a newborn with a fever and chills inside, supposedly to “cure a cold,” and then tried to stop the father from rescuing the baby.

The company has also invested hundreds of billions of dollars in developing what CEO Mark Zuckerberg calls “superintelligence.” Yet the evidence here points to no comparable effort to stop this material spreading across Meta’s platforms.

I think the gap is the story: Meta’s public case for its AI ambition sits beside a basic failure to act on accounts reporters have flagged. The announcement is quiet about what detection systems the company uses, whether anyone is accountable for repeat removals that do not stick, and what response time it considers acceptable.

Last month, Wired reported that Meta had allowed hundreds of ads featuring sexually explicit images of real children. Meta allows users as young as 13 on its platforms. Against that backdrop, leaving reported accounts live for more than a day looks less like a hard edge case than a test of whether the company still intends to moderate this material at all.