Meta’s AI wrongly deleted user accounts, NYT finds

Meta ↗ 0.84% used automated systems to delete Facebook and Instagram accounts that users had spent years building, and when those users tried to appeal, their cases were often handled by another AI rather than a human reviewer, according to a New York Times investigation published Monday.

The report found that Meta restored some banned accounts — including one with nearly a million followers — only after Times journalists intervened on behalf of the affected users. In some cases, the company told users their information would be "irretrievably deleted" and that no further review was possible.techcrunch+2

A Pattern of Automated Enforcement

Meta told the Times that the accounts examined in its reporting were banned by older moderation systems, not its latest AI tools. The company said its newer automated systems make 13 percent fewer mistakes and catch 10 percent more violations than human reviewers.threads+1

But for users caught in the system, that distinction offers little comfort. Business owners and creators reported losing accounts without being told what specific rule they had violated, and found that appeals were circular — routed back to automated processes rather than to human staff capable of exercising judgment.aiweekly

The problem is not new but appears to have intensified. In May, users worldwide reported a wave of mistaken deactivations across Instagram and Facebook. In January, more than 50,000 users said Meta had wrongly shut down their accounts over accusations of "heinous" behavior.reddit+2

Oversight Board Flags "Systemic" Concerns

The findings echo warnings issued weeks earlier by Meta's own Oversight Board, which in June said the company's account deactivation process raises "systemic human rights concerns" and a "lack of transparency and consistency." The board recommended Meta create a user-facing dashboard showing past violations, specific rules broken, and clear appeal options — none of which currently exist.mediapost+1

The board also called on Meta to disclose the role AI plays in content review and the imposition of penalties, noting that users often have no idea whether a human or machine made the decision that erased their digital presence.techcrunch

Fewer Humans, More Automation

The controversy arrives as Meta has sharply reduced its human workforce. The company laid off 8,000 employees in May and reassigned 7,000 more to AI projects, continuing a transformation that has left fewer people available to review moderation decisions. For affected users, the result is a system where machines make consequential choices and other machines adjudicate the complaints — with no human in the loop at all.nytimes+1