This study examines how armed conflict amplifies the perceived value of Generative AI (GenAI) while simultaneously eroding the conditions needed for its responsible deployment. Using Sudan's Emergency Response Rooms (ERRs) as a field site, we applied a two‑stage empirical approach: an exploratory survey of 89 community‑based humanitarian responders followed by 15 semi‑structured interviews with both GenAI users and non‑users.\ \ Findings indicate that war reshapes GenAI's role—its outputs become most valuable when information is hard to verify or obtain. Conflict makes on‑the‑ground expertise difficult or dangerous to access, prompting some volunteers to turn to GenAI for rapid answers. Yet participants consistently warned that GenAI cannot autonomously track rapidly shifting local conditions; relying on its results without human verification can lead to harmful decisions.\ \ Risks emerge even before GenAI is employed: carrying a phone or accessing the internet may expose volunteers to danger; sharing sensitive beneficiary or operational data with GenAI could jeopardize the safety of affected populations. Some respondents feared that AI‑augmented reports might undermine credibility with donors, whose funding sustains the food and medical aid provided by ERRs.\ \ Access to GenAI was uneven. Volunteers who could use it often depended on colleagues with relevant expertise or local knowledge to cross‑check its outputs. Our results suggest that responsible GenAI use during conflict depends not only on model performance or individual adoption, but also on whether responders can safely access the tools, verify outputs against on‑the‑ground realities, and recognize when limiting or avoiding GenAI is the more responsible choice.\ \ Review: The study highlights that in highly volatile settings, the benefits and hazards of advanced AI tools coexist. Only when secure access, reliable verification, and ethical vigilance are aligned can GenAI become a genuine asset for humanitarian action rather than a latent risk.