BSI commissioned Alma Economics to develop the economic modelling component of its report, The Cost of AI Failure: An Economic Study, which examines the potential financial risks associated with AI failures. This work formed part of BSI's wider programme of research on AI governance.

Our contribution focused on estimating the potential financial exposure associated with different types of AI failure, using real-world examples including biased or discriminatory decision-making, hallucinations and fabricated outputs, customer-facing chatbot failures, and agentic AI failures.

For each case study, we developed an economic modelling framework that considered the severity of the observed financial impact, the scale of the relevant market or economic activity, the extent of AI adoption, and, where relevant, the likelihood of the failure occurring. The estimates were then scaled to indicate potential financial exposure at a wider market level.

Our analysis suggested that AI failures could put approximately 0.1% to 3.6% of the relevant global market at risk each year, depending on the type of failure examined.

The final report combines Alma Economics' modelling with findings from BSI's survey of more than 1,250 business leaders, highlighting the importance of understanding and managing AI-related risks as AI adoption continues to expand.

Read the report here.