Meta's AI moderation is incorrectly suspending accounts.

      Camille Hanson was awakened one night in March by a notification: Meta had flagged her Facebook and Instagram accounts for deletion. She and her husband operated an English-teaching business on those platforms, boasting nearly a million followers. According to the New York Times, Meta claimed she had violated its policies on fraud and deception.

      She filed an appeal, but a week later, Meta rejected it. The response stated, “All your information will be permanently deleted. You cannot request another review of this decision.” Since the Hansons run their business from Portugal, her husband remarked that it felt like someone had closed their storefront overnight.

      Humans excluded from the process

      In March, Meta announced it would empower AI to determine which accounts violated its policies and manage the appeals. Months later, the company laid off thousands of employees, including those responsible for this particular work.

      Since then, over 60,000 users have signed a petition urging Meta to clarify its bans and to allow human reviews of appeals. Reddit forums have echoed similar grievances: automated systems suspend accounts without any personal response.

      Meta’s position

      Meta refutes the claim that AI is inferior to humans. Spokesman Daniel Roberts stated that its new moderation tools commit 13% fewer errors than human workers and detect 10% more violations. He noted that the five accounts highlighted by the Times were all banned using Meta's older tools.

      “We’re dedicated to reducing enforcement errors and assisting individuals in safeguarding their accounts, and AI is effectively achieving both,” he asserted.

      However, the company has had its challenges with automation. In May, hackers exploited its AI customer service chatbot against 34,000 Instagram accounts. Employees also protested a program that monitored their keystrokes to train the AI, and the AI advertising tools have encountered turmoil as well.

      A Kafkaesque situation

      The erroneous bans often carry severe accusations. Athenia Rodney, who has managed the group JuneteenthNY for 17 years, found her accounts deleted over claims of sharing child exploitation material. She contended that her content is family-friendly and that hackers had actually taken over her accounts.

      To recover them, she submitted identity documents, contacted the FTC, and messaged Meta staff via LinkedIn. Nothing was effective until the Times intervened. Meta reinstated one influencer only to ban him again days later for copyright violations, and then restored him once more.

      The European context

      In Europe, the stakes are heightened, as the same automated bans affect WhatsApp. Tech writer M.G. Siegler, who faced three bans, described his unexpected lockout from WhatsApp, without any warning or explanation. Many people across the continent rely on the app for daily activities, from work to childcare.

      Meta maintains that crucial decisions remain in human hands. After 26 former employees sued, asserting that an algorithm led to their layoffs, Meta stated these decisions are “made by people, not AI.” Yet, for millions of users, only a bot stands between them and account deletion. Appeals are also handled by a bot. The backlash is likely to continue as regulators intensify their scrutiny of platform accountability.