Why do AI models struggle with online hate speech detection?

On the International Day for Countering Hate Speech, the UN and UNESCO warn about the scale of the problem. A 2023 Ipsos-UNESCO survey shows over two-thirds of internet users encountered hate speech online. Meta, which owns Facebook and Instagram, has removed fewer hateful posts since 2023, with 1.3 million removals in the last quarter of 2025. AI systems tasked with detecting such content fall short compared to human judgment.