From Existential to Existing Risks of Generative AI - A Taxonomy of Who Is at Risk, What Risks Are Prevalent, and How They Arise

Explore a comprehensive research presentation that systematically analyzes 499 publicly reported incidents to construct taxonomies of Generative AI risks and their underlying sociotechnical failure modes. Learn how researchers from Carnegie Mellon University examined the expanding landscape of AI-related risks affecting people, communities, society, and the environment as Generative AI applications proliferate across diverse domains. Discover the methodology behind categorizing risks that arise during technology design, development, release, deployment, and downstream usage, with particular attention to privacy-related incidents. Understand the prevalence patterns of different risk types, failure modes, and affected human entities, including their co-occurrences in real-world scenarios. Examine findings that reveal how the majority of reported incidents stem from use-related issues while impacting parties beyond the direct end users of the faulty Generative AI systems. Gain insights into how tracing failure modes to their downstream real-world consequences provides actionable intelligence for policymakers, developers, and AI users, with emphasis on the critical need for prioritizing non-technical risk mitigation approaches in the rapidly evolving Generative AI landscape.