Arabic-speaking citizens in Israel routinely encounter government website documents written in Modern Standard Arabic (MSA), a variety that differs substantially from the Levantine dialect they use daily. This diglossia gap limits access to public services and creates informational inequity, especially for elderly citizens unfamiliar with modern language tools. We address this problem through three contributions. First, we present DAASI (Disglossia-Aware Arabic Simplification for Inclusion): a new parallel corpus of drawn from Israeli National Insurance Institute documents, constructed via GPT-4o draft simplification followed by expert human validation. Second, we conduct a sociolinguistic survey of 50 Arabic-speaking adults in Israel confirming that government documents are harder to comprehend than school-level texts. Third, we benchmark two Arabic transformer models and GPT-4o across multiple fine-tuning strategies and conduct human evaluations to assess simplified text. Our results confirm that existing Arabic NLP benchmarks do not generalize to the Israeli government domain and that DAASI fills a critical resource gap for inclusive Arabic NLP in a diglossic context.