Understanding AI Risks and Exposures
This course equips insurance professionals with a working taxonomy of AI failure. It shifts learners from being impressed by AI to recognizing how it breaks, identifying specific failure types, and understanding why AI risk correlates across portfolios.
Why AI risk is different Explain the difference between how conventional software fails and how AI fails. State why that difference makes AI failure harder to detect than an outage. Fabrication: the confident wrong answer Define fabrication and explain why fluency and accuracy are unrelated. Recognise the business situations where a fabricated output is most likely to cause loss. Bias and discrimination Explain the main routes by which bias enters an AI system. Explain why removing a protected characteristic from the data does not remove the bias. Describe why biased AI decisions create liability and regulatory exposure at scale. Opacity and over-reliance Explain what opacity means and why it is structural rather than concealment. Define automation bias and explain why expertise does not protect against it. Explain why a human in the process is not the same as a human in control. Drift Explain what drift is and why a model can degrade without anything changing inside it. Recognise the business conditions that accelerate drift. Data and intellectual property exposure Identify the points at which data is exposed by AI use. Distinguish the three intellectual property questions generative AI raises. Explain why ordinary staff behaviour is the most common route to AI data exposure. Third-party and embedded AI Explain why most organisations have AI exposure they never procured. Explain why asking an organisation whether it uses AI produces an unreliable answer. Adversarial AI Distinguish attacks on AI systems from attacks carried out using AI. Explain why AI raises the volume and the quality of social engineering at the same time. When the system acts Explain how an agent differs from an advisory system in risk terms. Explain why chained autonomous actions change the speed and scale of a failure. Concentration, aggregation and where the risk lands Explain why AI risk correlates across organisations that have no relationship with each other. Explain why correlation is different in kind from any individual failure. Locate a given AI exposure with the insured, their own firm, a third party, or across a portfolio.