This paper introduces the humanitarian passive as a machine-mediated syntactic pattern through which civilian suffering remains visible while responsibility becomes grammatically optional. Focusing on Palestine, Iran, and platform moderation, it defines responsibility loss as the measurable weakening of grammatical traceability between harm and responsible agency. The article proposes the Responsibility Loss Index (RLI) to evaluate whether AI-generated summaries, headlines, reports, and moderation notices preserve or erase agents responsible for violence, sanctions, restriction, censorship, or humanitarian harm. Its central contribution is to shift AI ethics from bias detection alone toward responsibility detection.