Meta’s Moderation AI Just Deleted Millions

Instagram and Facebook users woke up to deleted accounts this week. Some had been built over years. Some had tens of thousands of followers. All were gone. Then users tried to appeal. The appeals went to an AI. The same AI that deleted them.

Meta’s automated moderation made the call. Users contested it. And the system that decided they violated policy the first time got to decide if they were right to appeal it. That is not a guardrail. That is a loop.

The accounts got restored, but only after journalists started asking questions. Meta confirmed it after the pressure built. The company said it was fixing the issue. What it did not say was how many other accounts are still in that same automated loop, deleted by one AI, judged by the same system on appeal.

This is the part that breaks trust. It is not that an AI made a mistake. It is that the mistake went through a process designed to confirm itself. A human should have been in that loop somewhere. A different system should have reviewed it. Something should have broken the chain.

What worries me is this became news only because it hit accounts with enough followers that journalists noticed. How many smaller accounts are stuck in the same situation right now? How long will they stay deleted before someone at Meta decides to review them?

This is the cost of scale without friction. When moderation is fast and automated all the way through appeal, you get speed but you lose the thing that makes a process feel fair: a human somewhere who stops and thinks.