Artificial intelligence (AI) has experienced widespread adoption across diverse sectors due to its capacity to enhance operational efficiency and economic competitiveness. However, the deployment of AI systems has also simultaneously introduced numerous security challenges and potential risks which demand careful consideration. As investments in AI development have increased substantially, corresponding investments in cybersecurity have become more critical. Ensuring the secure implementation of AI, particularly within critical infrastructure systems, necessitates the development of robust and resilient systems. Analysing real world AI incidents provides valuable insights which may serve to enhance security mechanisms and prevent future vulnerabilities. Since the inception of artificial intelligence, researchers have identified various system vulnerabilities and associated risks. Promoting awareness of such potential AI hazards has proven instrumental in facilitating a deeper understanding of both the scope and severity of these risks. Moreover, such awareness provides a framework for developing AI tools which are not only more resilient, but also ethically sound. It is essential to examine how these dynamics evolve in practical environments by systematically identifying such incidents, their underlying causes, and their consequent impacts, rather than relying solely on theoretical projections. In response to this urgent need, AI incident databases have emerged as crucial instruments for responsible AI development and governance. The primary objective of these initiatives is to methodically document and categorise incidents, thereby strengthening security measures, informing preventative strategies, and fostering transparency and accountability within AI systems management.
Ana Kovačević (Wed,) studied this question.