A research group has found that algorithms struggle to differentiate anti-Asian violence-provoking speech from general hate speech, leading to a potential increase in real-world attacks. They call for internet and social media moderators to strengthen their detection and intervention protocols.