Researchers at the Center for Countering Digital Hate trained an AI model to flag Facebook comments that would breach Meta’s rules on violence, hate and harassment. After the company eased those policies, violent‑threat comments jumped from 1,800 to 7,600 in a six‑month span, while hate‑speech posts rose from 6,900 to 30,000. Overall the system flagged about 60,000 rule‑breaking posts.

Meta’s spokesperson told WIRED the firm regularly publishes violation metrics and saw no rise in hateful conduct through 2025, but declined to comment without the full study. WIRED released a sample of the abusive comments hours before they vanished from Facebook. The company’s 2025 transparency data shows proactive enforcement fell roughly 50% after the rule change, and the cut also slowed response times for reported content.

Senator John Curtis warned that trimming oversight on violence, hate and harassment inevitably fuels abuse, a point echoed by the CCDH authors who noted the rise in toxic posts mirrors the drop in enforcement. Analysts say this pattern could influence upcoming regulatory hearings. The correlation suggests Meta’s policy shift directly weakened user safety, giving regulators a data‑driven case for stricter moderation.