Building trust in AI during crises: Why ethical design matters

From wildfires to floods, artificial intelligence (AI) is transforming how organizations communicate during emergencies and disasters. But as machines help generate alerts and safety messages, one big question remains: Can people truly trust what AI says during a crisis?

That question guided our international research team of scholars and crisis communication professionals. Supported by the Page Center, a team of scholars (Jaekuk Lee, Yan Jin, Wenqing Zhao and Nicole Cortes) and I joined Philippe Borremans, risk, managing director of RiskComms, in exploring how AI-generated messages shape public trust and professional responsibility during crises.

Our project surveyed both crisis communication practitioners and the general public in the United States and the United Kingdom. We wanted to understand, in the context of emergencies and disasters:

What we found

For practitioners, AI brought both promise and pressure. U.S. professionals reported a higher level of perceived susceptibility of negative consequences of adopting AI tools in the emergency and disaster communication management practice compared to U.K. professionals. U.K. professionals reported a greater intention to experiment with AI in emergency and disaster communication management practice compared to U.S. professionals. Among both U.S. and U.K. professionals, individual-level READINESS regarding the adoption of AI, including adaptability and self-efficacy, serves as an important predictor of their perceived accountability of addressing AI-triggered threats and their intention to be transparent and disclose their use of AI.

Interestingly, those who used AI more frequently also reported higher anxiety regarding being replaced by AI and perceived risks (perceived severity and susceptibility), they worried about potential errors (e.g., privacy violation) when public safety was at stake.

When we turned to the public, participants valued transparency. They wanted to know when AI was involved in creating a message. When that disclosure was made clear, trust increased, especially when messages highlighted human oversight, what we call a “human-in-the-loop” approach.In the U.K., higher public trust in AI led to stronger organizational trust. In the U.S., it encouraged word-of-mouth advocacy, people were more likely to share messages with others.

Why It matters

AI isn’t just a new tool — it’s a new communicator. The way organizations design, disclose, and deploy AI systems directly affects public confidence, compliance, and credibility.To guide ethical AI use, our team developed a Human-in-the-Loop Ethical Protocol — a framework that ensures every AI-generated message is reviewed by a human expert for accuracy, ethics, and empathy. The goal: Combine AI’s speed with human accountability.

Collaboration across practice and academia

This project shows the power of scholar–practitioner collaboration. Communication professionals from the International Association of Risk and Crisis Communication (IARCC) helped refine survey questions to reflect real-world concerns like data privacy and message reliability.

Graduate and undergraduate students from North Carolina State University and the University of Georgia were vital to the project, supporting data collection, monitoring, and now participating in analysis and publication. Their involvement bridges research, education, and professional practice.

For more information about this study, email Cheng at ycheng20@ncsu.edu. This project was supported by a 2024 Page/Johnson Legacy Scholar Grant from the Arthur W. Page Center.