Model validation
Agentic AI set to drive end-to-end automation of MRM workflows
Risk Live: Validator role likely to shift towards oversight and expert judgement
Managing AI models is reshaping three lines of defence, say banks
Risk Live: Model managers want seat at table during development, and expect first line to take charge of AI model testing
How banks are using AI assistants for credit risk model validation
Model validation has become a structural bottleneck for European banks. AI-based assistants can offer a credible response and early production deployments are having an operational impact
Regulators question human-in-the-loop as AI governance tool
Bank of England and FSB executives suggest it’s more important to retain overall accountability
Do banks still need to validate GenAI models?
Regulators carved out GenAI models from new risk guidance. Banks shouldn’t see this as a reason to stop validating them.
Hopes, fears and ‘mass confusion’: the sudden end of SR 11-7
Banks welcome chance to prioritise model reviews, but fret over future policy changes and AI
Graph neural networks for credit default prediction: robustness and model evaluation
The authors evaluate the robustness and performance of graph-based models in credit default prediction.
AI risk management and the shift to capability control
By reframing validation, banks can align innovation with regulatory demands and maintain robust risk discipline, argues risk manager
The do-it-all machine: model risk in the age of generative AI
Banks race to understand risks posed by new breed of multi-purpose bots
Model validation of a generative-artificial-intelligence-based avatar for customer support in banking
The authors put forward a validation method for a gen-AI-based avatar designed to deal with customer inquiries in the banking sector.
Rising reliance on internal auditors spooks regulators and industry
Risk managers warn US is substituting supervisors with auditors; could compromise independence
Generative artificial intelligence in model risk management: emerging opportunities, supervisory challenges and validation frameworks
The author proposes a structured approach to validating generative AI models in line with the principles of current regulatory standards.
The loneliness of the model risk manager
Boards may see them as a drag on innovation; risk functions need to show they embrace efficiency
A dual backtesting framework for quantifying nested model error and unlocking capital efficiency
The author puts forward a framework for dual backtesting, in which single-blind backtesting assesses core models and double-blind backtesting evaluates the whole system.
FHLB Cincinnati explores AI to spot failing banks
Agentic model detects anomalies, monitors sentiment and drafts credit reports for analyst review
Exceedance-based backtesting of expected shortfall
The authors apply exceedance-based validation techniques often used for VaR model validation the the validation of ES models, showing such an application to be feasible.
Rethinking model validation for GenAI governance
A US model risk leader outlines how banks can recalibrate existing supervisory standards
Clearing houses warn Esma margin rules will stifle innovation
Changes in model confidence levels could still trip supervisory threshold even after relaxation in final RTS
Interpretable machine learning for default risk prediction in stress testing
This paper proposes a benchmark model which can be used to predict the forward-looking probability of default of a real-world credit card portfolio.
Statistically distinguishable rating scales
The author suggests a means to design a statistically distinguishable rating scale that is not excessive in relation to the existing observation statistics, allowing for more stable validation.
Want to be a quant? Here’s how (and how not) to get hired
Stay curious, be a team player, speak well – and don’t be big-headed