Hate rarely starts with slurs. It starts with stories, carried in everyday words, that repeat until a fringe idea begins to sound like common sense. These stories travel from fringe forums to mainstream social media and into the AI systems that learn from everything online, which then repeat them back to millions of people.
Moderation removes hate after it appears. AddressHate works earlier, on the stories that make hate feel normal. We measure what existing tools miss, the coded language and hidden meanings beyond slurs, and show which platforms and AI systems amplify it. We fund the research, publish the standards, and build the tools that let regulators, researchers, advocates, and companies spot hateful narratives while they’re still spreading, and stop them before they become real-world harm.
AddressHate is a 501(c)(3) nonprofit working to break that cycle, so the next hateful narrative is seen while there’s still time to act.
We fund academic researchers to build hate taxonomies that define how hate is expressed, including its coded forms. In internal testing, our detection model identified implicit hate with the same accuracy as leading frontier AI models (e.g., ChatGPT Astra 6, Claude Opus 5.5) at about 1% of the cost. Digital Hate Review, the first peer-reviewed journal on digital hate, brings current research from leading labs and motivated students to advocates across fields.
Those guides become the foundation for labeling online content, which informs our detection work. The AddressHate Annotator lets researchers label content twice as fast as comparable tools and is already in use with six research partners.
Our dashboards show advocates and researchers which hateful stories are spreading, where and how fast. Public scorecards reveal how much coded hate major platforms and AI models miss. And our benchmarks, standard test sets labeled by experts against a peer-reviewed taxonomy, let regulators, auditors and researchers verify for themselves what platforms detect and what they don’t.
We partner with tech companies, AI labs, trust and safety teams, educators, community advocates, and security teams to turn our intelligence and research into action: free safety tools for smaller platforms, early warnings for advocates, and classroom materials that prebunk hateful stories before they take hold.