Meta hate speech removals drop 79% following reactive moderation pivot
Officials and experts at a Royal Commission discussed content moderation challenges, including Meta's move to a reactive moderation model in early 2025 to prioritize accuracy. The streaming platform Kick faced scrutiny regarding its reliance on an outsourced Serbian moderation team and high rates of flagged content labeled as false alarms.
Key Takeaways
- Hate speech enforcement actions on Facebook fell from 5.8 million in Q4 2024 to 1.2 million by Q3 2025.
- Streaming site Kick reported that just 0.98% of 179,914 moderation reports resulted in action between January and May 2026.
- Meta's internal policies allow statements like "White people are all Nazis" and "Trans people don’t exist" to prevent policing general offensiveness.
- Kick outsources its entire moderation workforce to a team of 136 employees based in Serbia.
- Coded AI-generated content, such as "My Little Pony" Holocaust denial memes, is bypassing standard algorithmic detection.
Why It Matters
The retreat from proactive moderation marks a significant reversal in trust and safety strategy for major social platforms. By shifting the burden of identification onto users to avoid censorship claims, platforms are seeing a massive drop in actionable enforcement. For the streaming ecosystem, this highlights a widening gap between automated detection and evolving "coded" hate speech that uses AI-generated imagery and linguistic workarounds. As platforms like Kick rely on lean, outsourced human teams and Meta scales back its fact-checking architecture, the industry faces increasing regulatory pressure to define where "free expression" ends and platform liability for hosted content begins. Watch for whether Australia’s eSafety Commissioner issues formal notices to Kick following these revealatory enforcement statistics.
Additional Context
The Royal Commission's findings arrive as Meta faces intense scrutiny for its 2025 policy overhaul, which reportedly included the removal of key protections for LGBTQ+ users and the termination of various Diversity, Equity, and Inclusion (DEI) programs. Per QNews in July 2026, these changes were specifically designed to "allow more speech" on the platform following political shifts in the U.S., including the re-election of Donald Trump. This pivot has drawn sharp criticism from advocacy groups like GLAAD, which argues that the reduction in proactive enforcement has made platforms fundamentally less safe for marginalized communities. Simultaneously, the streaming platform Kick has attempted to formalize its lenient approach through a March 2026 update to its Community Guidelines. According to Streams Charts, Kick consolidated its rules from 14 to 11 sections and introduced a "Context and Intent" clause. This framework mandates that moderators consider whether a violation was accidental or if the streamer took proactive measures to mitigate harm. While the platform permits AI-driven content with strict disclosure, its general counsel confirmed to the Royal Commission that Kick has no formal engagement with experts or organizations focused on combating antisemitism. Meta also recently expanded its hate speech policies to crack down on the term "Zionist" when used as a proxy for Jewish people in dehumanizing contexts. Per The Guardian in July 2026, Meta's director of content policy acknowledged that while overall enforcement has dropped, the company is attempting to focus on content that causes offline harm. However, the 79% drop in actioned items on Facebook—and a similar reduction on Instagram from 7.4 million to 2 million items—suggests the reactive model is struggling to maintain the same volume of moderation as previous AI-driven proactive systems.
Read full article at heraldsun.com.au
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source