Camille Hanson and her husband, Calvin, were surprised when Meta, the parent company of Facebook and Instagram, flagged and banned their accounts earlier this year, citing violations related to fraud and deception. The Hansons, based in Portugal, run a business teaching English to nonnative speakers and collectively manage nearly one million followers on their social media accounts. The notifications came without clear explanations, and an initial appeal filed by Ms. Hanson was denied, with Meta stating the decision was final and their data would be permanently deleted.

The incident highlights growing concerns over Meta’s increased reliance on artificial intelligence (AI) for content moderation and account enforcement decisions. Earlier this year, Meta announced plans to expand the use of AI tools to determine which accounts to ban and to remove harmful content. This shift was accompanied by large-scale layoffs, including employees previously responsible for such moderation tasks, leaving users with limited access to human review or customer support.

Online communities have since mobilized, with more than 60,000 users signing a petition calling for greater transparency and human oversight in Meta’s enforcement processes. Meta, however, maintains that its AI systems reduce errors, citing internal data that shows the new AI tools make 13 percent fewer mistakes than human moderators while detecting 10 percent more violations. A Meta spokesperson also noted that the company’s AI banned all five accounts flagged for review by a news organization, asserting confidence in the technology’s effectiveness.

Despite these assertions, examples continue to emerge of AI missteps affecting users. Athenia Rodney, founder of JuneteenthNY, a group commemorating the end of slavery in the United States, had all her personal and business accounts banned last year for alleged violations without detailed explanations. Ms. Rodney said the accounts had recently experienced suspicious login attempts, suggesting they had been hacked prior to the bans. Despite multiple efforts to recover her accounts—including submissions of identification, complaints to consumer protection agencies, and direct outreach to Meta employees—her access remained restricted for months, causing damage to the group’s reputation and sponsorship relationships.

Following renewed scrutiny, Meta revisited five flagged accounts, determining that two had been wrongfully banned, including the Hansons’. The company restored those accounts and apologized for the error. Another notable case involves Roberto Suarez, an Instagram influencer with more than 500,000 followers. After being initially banned, his account was restored, only to be suspended again weeks later over alleged copyright violations, which Mr. Suarez denied. The account was ultimately reinstated once more.

Meta’s AI moderation systems have also faced challenges beyond user bans. In May, hackers utilized Meta’s AI customer service chatbot to compromise tens of thousands of Instagram accounts, including that of former U.S. President Barack Obama. Earlier that month, Meta paused a controversial employee monitoring program designed to gather data for AI training in response to internal backlash. More recently, a lawsuit by 26 former employees alleges that Meta used AI algorithms to disproportionately target and lay off individuals with disabilities—a claim Meta denies, stating human decision-making guided workforce management.

As Meta continues to expand its AI-driven policies, the experiences of users like the Hansons and Ms. Rodney underscore broader concerns about the balance between automated enforcement and the availability of human review mechanisms. The company stresses its commitment to improving accuracy and user protection while navigating the complexities of content moderation at scale.