UK FCA Opens Consultation on AI Model Governance for Financial Services Firms

The UK Financial Conduct Authority opened a consultation on AI model governance requirements for regulated financial services firms, proposing mandatory explainability documentation, periodic bias audits, and human override mechanisms for all models making consumer-facing decisions.

The proposed rules would apply to AI systems used in credit decisioning, insurance pricing, fraud detection, and investment recommendations. Firms would be required to maintain a "model registry" documenting each AI system's purpose, training data sources, known limitations, and performance metrics.

Bias audits would be required annually for models affecting protected characteristics (age, gender, ethnicity, disability). Firms that cannot demonstrate statistical fairness across protected groups would face enforcement action, regardless of whether discriminatory intent existed.

"AI will transform financial services for the better — but only if consumers can trust that algorithmic decisions are fair and explainable," said FCA CEO Nikhil Rathi. "We are not regulating the technology. We are regulating the outcomes."

Industry groups expressed measured support for the principles but raised concerns about the compliance burden on smaller firms. The consultation proposes a proportionality threshold: firms using fewer than five AI models with annual decision volumes under 100,000 would face lighter requirements.