Meta's AI moderation is deleting long-running accounts by mistake

Imagine losing a 10-year-old account with your whole business in it, and the “judge” reviewing your appeal is the same AI that banned you.

What the Times found

That is exactly what is happening at Meta right now. The New York Times reported on July 21, 2026, that Meta’s automated moderation has been mistakenly deleting Instagram and Facebook accounts that people spent years building, including business accounts with real customers behind them.

Users describe a specific pattern: bans that arrive without warning, without explanation, and without a human anywhere in the process. One day the account with a decade of posts, followers and customer relationships exists; the next day it does not, and no one at the company will say why.

The appeal loop

The worst part is what happens next. Appeals frequently get processed by the same automated system that made the original mistake. People stuck in the loop say there is simply no human to reach: the error and the review of the error run on the same machinery.

Meta’s statistical defense

Meta does not deny using AI at scale. Its defense is statistical: the company says its newer AI tools make 13 percent fewer moderation mistakes than human reviewers and catch 10 percent more violations.

Both numbers may be true, and both are cold comfort if you are the false positive. Aggregate accuracy is the platform’s metric; the individual account holder experiences only the binary outcome, and at Meta’s scale even a small error rate translates into an enormous number of wrongly banned people.

As platforms push more moderation to AI, the operating principle quietly shifts from “innocent until proven guilty” to “banned until the bot changes its mind.” For anyone whose livelihood runs through an Instagram or Facebook account, that shift is not abstract. The next purge could be closer than it looks.