NUGGETWHAT WOULD YOU DO?

What would you have done? — Meta let its AI ban accounts, and the appeal was judged by the same AI

November 16, 2026

In 2026, Meta increasingly let its artificial intelligence suspend the Facebook and Instagram accounts that break its rules, at massive scale. Faced with the volume, three options were on the table: (A) automate both the penalty and the review of appeals, with very little human review; (B) let AI detect and flag, but route every account suspension and every appeal to a person; (C) keep human moderators for any decision that takes down a whole account.

Meta chose (A). Perfectly legitimate accounts were disabled by mistake, small businesses among them: a business erased after fifteen years, a photographer locked out for eight months. The problem wasn’t the error, which is unavoidable at scale, but the lack of recourse: the appeal was often routed back to the same AI and denied within minutes, with no way to reach a person, and some accounts were restored only after journalists stepped in. Automating a suggestion and automating a penalty are not the same decision: a sanction no one can review becomes an irreversible error that lands on innocent people. Before letting an AI refuse, suspend, or exclude a customer, a supplier, or an employee, put the recourse path in writing: a named person must be able to review and reverse the decision, never the same AI that made it.

Source: NBC 5 Investigates and The New York Times, investigations into account suspensions by Meta’s AI moderation (April-July 2026).

Next step

Does this finding apply to your business?

An AI diagnostic checks it against your answers and documents

← All nuggets

Read-only snapshot of the Institute's published content as of 2026-10-14. Nuggets refreshed on 2026-11-22.