Published
2026-04-24
Conference Proceedings Volume
Section
Artificial intelligence, digital tools, and innovative technologies for crisis management
License
Copyright (c) 2026 Sandra Ćavar, Ivan Ćavar, Antonio Krsnik
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
How to Cite
Risks of Disinformation in Crisis Management: What Happens When AI Processes Fabricated Inputs?
Sandra Ćavar
Ivan Ćavar
Antonio Krsnik
Keywords: artificial intelligence, crisis management, disinformation, LLM, data verification, decision-making
Abstract
In crisis situations, decisions are made quickly and under pressure. What if an AI system implemented in a crisis management center processes unverified or fabricated inputs? This paper explores how generative artificial intelligence responds to falsified crisis reports, inspired by the "bixonimania" case – a fictional disease mistakenly treated as real by ChatGPT and Gemini. Through a series of simulated reports with invented locations and data, the study analyzes recurring patterns of flawed reasoning, including the generation of priority actions without adequate source verification. The analysis focuses on recognizing risks of automated decision-making in crisis management centers. The paper proposes a framework for safe integration of AI systems that includes verification protocols, limitations on autonomy, and the preservation of essential human responsibility in critical decision processes.
References
Osmanović-Thunström, A. (2026). "Izmislili bolest kako bi nasamarili AI”. Bug.hr. https://www.bug.hr/umjetna-inteligencija/izmislili-bolest-kako-bi-nasamarili-ai--i-uspjelo-im-je-58878
Scientific Advice Mechanism (2025). "How far should we trust AI in a crisis?"
https://scientificadvice.eu/news/how-far-should-we-trust-ai-in-a-crisis-european-experts-weigh-in/
EU AI Act (2024). "Article 6: Classification Rules for High-Risk AI Systems". https://artificialintelligenceact.eu/article/6/
Scientific Advice Mechanism (2025). "AI in crisis management".
https://scientificadvice.eu/advice/ai-in-crisis-management/
European Commission (2025). "Artificial intelligence has potential to improve emergency and crisis management".
