The Machine Shows the Victims, But Hides Who Caused the Suffering

AI can describe war, sanctions, censorship, and humanitarian crisis without lying. The danger is subtler: it can keep the suffering visible while removing the responsible actor from the sentence. That pattern is called the humanitarian passive. Example: “Military forces bombed a residential building.” becomes: “A residential building was damaged amid escalating violence.” The harm remains.The agent disappears. The paper calls this effect responsibility loss: the measurable weakening of the grammatical link between harm and responsible agency. Palestine shows this through direct violence, occupation, displacement, humanitarian crisis, and platform moderation. Iran shows it through sanctions, banking restrictions, overcompliance, medicine shortages, military pressure, and security framing. The core claim: Suffering remains visible. Responsibility disappears. AI ethics should not only detect bias, hate speech, misinformation, or toxicity. It must also detect whether machine-generated language preserves the grammar needed to name responsibility.