Meta Sees Rise in Online Harassment but Fewer False Flags After Moderation Overhaul

Meta, the parent company of Facebook, Instagram, and Threads, has reported a modest increase in online harassment and graphic content during the first quarter of 2025. The uptick follows sweeping changes to its content moderation policies earlier this year—changes designed to reduce enforcement mistakes and allow for more open political discussion across its platforms.

In its Q1 2025 Community Standards Enforcement Report released Thursday, Meta revealed that bullying and harassment content on Facebook rose from 0.06–0.07% in Q4 2024 to 0.07–0.08% in Q1 2025. The company attributed the rise to a spike in the sharing of policy-violating content in March.

The prevalence of violent and graphic content also climbed slightly—from 0.06–0.07% to approximately 0.09%. Meta said the increase was partly due to users sharing more violating content and partly due to its efforts to cut down on enforcement errors.

In a statement to Business Insider, Meta spokesperson Erica Sackin clarified that while the percentage change may seem minor, it represents a real difference in user experience: “What 0.06% to 0.07% actually means is that prevalence of that type of content went from 6 views out of every 10,000 to 7 views out of every 10,000.”

Policy Shift: Less Strict, More Political Speech

Much of the change can be traced back to January, when Meta implemented a major overhaul of its content moderation strategy. The company relaxed restrictions on several politically sensitive topics—including immigration and gender identity—and narrowed its definition of hate speech. Instead of policing a broad range of inflammatory statements, Meta now focuses on “direct attacks” and “dehumanizing speech.”

Previously, content could be flagged for expressing contempt or disgust, using curse words, or calling for exclusion. Those criteria have largely been removed, with Meta arguing that such content may be distasteful but is also a legitimate part of political discourse.

The biggest structural shift came when Meta eliminated third-party fact-checkers and introduced a crowd-sourced “Community Notes” feature similar to the one used on X (formerly Twitter). These community-driven annotations aim to provide context on questionable posts, but some experts have warned that relying on user-generated corrections could open the door to bias and manipulation.

Lower Error Rates, But Not Without Tradeoffs

Despite the rise in problematic content, Meta emphasized that its new approach is working as intended—at least in some areas. Under its previous moderation model, the company removed millions of posts daily, but estimated that 10% to 20% of those takedowns were mistakes. Since revamping its policies, Meta claims to have cut those false positives in half.

“Our Q1 2025 report is the first quarter where these changes are reflected in the data,” said Sackin. “Across a range of policy areas, we saw a decrease in the amount of content actioned and a decrease in the percent of content we took action on before a user reported it.”

The company says it’s now aiming to “strike the right balance” between over-policing its platforms and allowing harmful content to spread. That sentiment was echoed by cybersecurity expert Theresa Payton, who told Business Insider that while crowd-sourced fact-checking represents a “democratization of content moderation,” it also carries risk.

“You have to end up having both technology monitoring community notes to make sure you don’t have bots or people who don’t have pure intentions,” Payton said, “and there could be an opportunity for bias or misinformation there as well.”

Historical Context and What Comes Next

The recent rise in violations may raise eyebrows, but it’s consistent with historical trends. Meta’s own data shows that harassment and graphic content often peak in the second quarter of the year before tapering off. In both 2023 and 2024, harassment violations hovered around 0.07% in Q4 but hit 0.08%–0.09% in Q2.

Whether the increase in Q1 2025 is part of a familiar cycle or an early indicator of deeper problems remains to be seen. For now, Meta is signaling that it’s willing to accept a modest rise in harmful content if it means fewer errors and more space for debate.

Still, the company will be closely watched in the months ahead as it tries to maintain user trust while walking a tightrope between free expression and online safety.