Bank of England AI Consortium Sounds Alarm on GenAI Risks
The Bank of England AI Consortium has sounded the alarm on the limitations of traditional model risk management frameworks as generative artificial intelligence (GenAI) adoption accelerates in UK and US financial services.
Legacy model validation methods, such as those outlined in the Prudential Regulation Authority's (PRA) SS1/23 Supervisory Statement on Model Risk Management, are no longer sufficient for GenAI models that combine multiple components and rely on third-party APIs.
The consortium has proposed a shift towards system-level risk oversight, agentic payment controls, and automated failure containment. This includes outcomes-based validation, whole-system governance, and standardised incident response protocols.