A new paper analyzes how online platforms map their terms and conditions to the Digital Services Act (DSA) Transparency Database taxonomy. Using a GPT-4.1 classifier on 447 policy documents from 31 services, the study found that over 99.8% of moderation decisions cited violations of the platform's own terms and conditions rather than legal provisions. The research also revealed that the applicability of high-level categories varies significantly by service type, and that the database's schema design may contribute to the prevalence of catch-all reporting. AI
IMPACT This research highlights how LLMs can be used to analyze policy documents and suggests improvements to regulatory schemas for better clarity.
RANK_REASON The cluster contains an academic paper analyzing policy and LLM application. [lever_c_demoted from research: ic=1 ai=0.7]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →