Open Problems in AI Incident Governance

AI systems may produce failures after deployment that pre-deployment safety assessments do not anticipate. Managing these failures requires what we refer to as adequate AI incident governance, where having good definitions, taxonomies, monitoring practices, reporting mechanisms, and incident analysis is essential. We examine existing frameworks related to AI incident governance by regulatory bodies and independent efforts, and find that while there are frameworks that describe how individual functions can be performed, there is a lack of consistency within the aspects of definitions, classification, monitoring, and reporting. These inconsistencies apply to the types of incident data that is collected and reported, the ways in which they are categorised, and as a result, the depth, representativeness, and accuracy of analysis that can be performed. We identify open problems at each stage of the incident governance pipeline, and find that the absence of standardised monitoring and reporting requirements constitutes a significant gap. To address this, we propose a set of principles supported by concrete monitoring guidelines and a reporting template to facilitate their implementation.