Mapping the Generative AI Risk Space

Om kurset

Nick van der Meulen reads MIT CISR's January 2026 research briefing, which he co-authored with Hippolyte Lefebvre and Barb Wixom. See the text version and related content at https://cisr.mit.edu/publication/2026_0101_GenerativeAIRisk_VanderMeulenLefebvreWixomLegner. Abstract: Generative AI (GenAI) risks have multiplied as organizations progress from GenAI tools and solutions to AI agents—amplified by massive, decentralized demand. But not all risks are alike. This briefing maps the GenAI risk space, identifying which components of GenAI implementations give rise to risk: from training data and foundation models through prompts, outputs, and use decisions. RAG and AI agents expand this space further. Drawing on 62 interviews with data and technology executives, we distinguish embedded risks—inherited with the technology—from enacted risks, which emerge from organizational choices. Each type of risk requires a different response. Leaders should inventory their GenAI deployments, assign accountability, and establish audit trails that log prompts, outputs, and human interventions.