Meta Runs Child Abuse Ads in India

You have seen the takedown notice. Every time a report like the one BBC Eye published on Meta’s advertising in India forces the removal of an unlawful advertisement, the company can point to the deletion as proof that its systems work. You have not yet looked for the account that opens the next morning under a different name, offering the same wares to the same buyers, because the thing that was destroyed was the window, not the man who kept throwing the stone.

I do not say this to excuse what is alleged. If it is true that ads promoting the abuse of children have continued to run on Meta’s platforms in India, the guilt belongs first to whoever placed them, and second to whoever built a machine so well-tuned to find buyers that it could not distinguish a customer for shoes from a customer for something no civilized ledger should record. But let us follow the money a little further, because the report itself, and the credit Meta receives for acting on it, are both symptoms of the same design.

Consider what the advertising engine is built to do. It is built to find, among a billion strangers, the ones most likely to respond, and it is rewarded, in revenue and in engineering praise, for doing this with extraordinary precision. Nobody at Meta wrote a rule that says “find buyers of exploitation material.” But the machine was told to find buyers of whatever is offered, and it obeyed with the same diligence it brings to selling sandals. The visible achievement, the thing shareholders applaud, is the matching. The invisible cost is that the matching does not know, and in a sense cannot know without being told by a journalist, what it has matched.

Now watch what happens after the takedown. The company earns its second round of applause for removing the ad once exposed. But removal after exposure is not prevention; it is confession dressed as correction. The unseen victim here is not only the child harmed before anyone noticed, which is grave enough. It is the next child, whose exposure depends entirely on whether a BBC reporter, and not Meta’s own systems, happens to go looking. A safety mechanism that activates only under the gaze of outside investigators is not a safety mechanism. It is a fire alarm that rings only when a stranger walks by with a match already lit, and calls this vigilance.

I am told the report is an allegation, not yet a settled fact, and I take that seriously; a man who counts unseen costs owes the same rigor to unseen exaggerations. But notice that the incentive I describe does not depend on whether this particular claim proves true. It depends only on how the advertising engine is built, and that engine’s structure is not contested by anyone, least of all Meta.

So the real question the coverage skips is not whether the ad was seen and removed. It is how many were never seen at all, and what it would cost, in engineering hours and advertising revenue foregone, to build a system that looks for such buyers before a reporter has to.