Addressing AI Risks in Critical Infrastructure: Formalising the AI Incident Reporting Process

Integrating Artificial Intelligence (AI) into critical infrastructure sectors such as the power grid, energy systems, and telecommunications has revolutionised decision-making processes, enhancing efficiency and reliability. However, this data-driven transition introduces inherent risks, including adversarial attacks, biases, and system failures with potentially catastrophic consequences. This paper addresses the crucial need of collecting AI-incident data to develop effective mitigation methodologies. It analyses the successful incident reporting mechanisms in safety-critical sectors of aviation, cybersecurity, and fire safety, which have significantly improved system reliability and prevented future incidents. It then evaluates the existing AI incident databases, examining reported incidents, impacted sectors, and affected parties. Findings of this study reveal significant gaps in capturing critical infrastructure-specific AI incidents. The paper proposes a formalised process for AI-incident reporting for critical infrastructure, enabling data collection and analysis, thus assisting in developing mitigation strategies, safety protocols, and regulatory frameworks. Emphasising transparency, accountability, and collaboration, this study ensures responsible and trustworthy deployment of AI in critical infrastructure.