Tuesday 17 March 2026
08:00 - 08:50
Registration and breakfast
08:50 - 09:00
Chair’s Opening Remarks
Day 1 Chair:
09:00 - 09:40
THE FUTURE OF MODEL RISK IN THE AI ERA
Opening Keynote: How GenAI and agentic AI are transforming model risk management reshaping validation, ethics, and the role of MRM as an innovation partner
- Reviewing how GenAI and agentic AI are entering risk models and decision-making tools in banks
- Examining model risk implications when using AI: transparency, explainability, fairness, bias, ethical concerns
- Discussing evolving role of the MRM function: from validator to enabling partner in AI innovation
- Exploring challenges of validating black-box AI models: identifying features, documenting behavior, monitoring drift and unintended outcomes
09:40 - 10:10
View from the Regulator
Keynote presentation: Evolving Expectations for Model Risk Management in a Dynamic Regulatory Landscape
- Insights from the Bank of England’s approach to model governance and independent review
- Enhancing resilience and accountability across model lifecycle management
- Supervisory expectations for AI, machine learning, and advanced analytics models
- Future directions: integrating model risk into broader prudential and systemic risk frameworks
10:10 - 10:40
MODEL ECOSYSTEM
Session: Structuring effective AI governance involving risk, compliance, technology, and business teams
- Structuring multifunctional governance involving risk, IT/data, compliance, and business lines for AI model oversight
- Defining first-line vs second-line responsibilities in model development, validation and monitoring with AI usage
- Managing decentralised innovation (e.g., business line AI teams) vs centralised oversight (AI centres of excellence) and reconciling both
- Measuring governance effectiveness: escalation procedures, incident management, model inventory, audit readiness
10:40 - 11:10
Coffee and networking break
11:10 - 11:50
REGULATORY DIVERGENCE
Panel Discussion: Navigating global regulatory fragmentation and its implications for cross-border model governance
- Mapping and comparing regulatory regimes: EU AI Act, UK SS 1/23, US SR 11-7 (and equivalent) and their impact on model risk
- Understanding how banks with global footprint must adapt to divergent regulatory expectations and harmonisation gaps
- Discussing how regulatory burden in European countries is increasing and whether model risk functions are becoming compliance-driven rather than risk-focused
- Exploring how evolving frameworks influence validation, governance, documentation and audit expectations
- Examples of strategies for staying ahead: regulatory horizon scanning, embedding regulatory change into model risk lifecycles, building flexibility
Ratul Ahmed Group Head of Model Risk Management & Validation, Commerzbank
Karolos Korkas Head of Algorithmic Trading Model Risk, AI Model Risk & Compliance Model Risk, Nomura
Ying Poikonen Executive Director - Head of Modelling Group EMEA Region, SMBC
Suresh Sankaran Prudential Regulation Lead, NatWest
11:50 - 12:20
MODEL RISK MANAGEMENT 3.0
Session: From Ethical & Responsible AI/Gen AI to Model Risk Governance Intelligence: How the OECD AI Principles are reshaping the MRM Execution
- Analyze the OECD AI principles and their implication on model risk management practices
- Formulate an integrated MRM framework inclusive of AI/Gen AI models
- Explore metrics used for assessing AI/Gen AI models explainability, fairness, bias and safety
- Deep dive in the impact on model validation activities and model governance
- Real world applications and case studies
12:20 - 13:20
Networking Lunch
13:20 - 14:00
MODEL VS METHOD
Panel Discussion: Navigating grey zones in model classification to align validation scope, regulatory expectations, and governance consistency
- Clarifying definitions: what constitutes a “model” vs a “method” or “tool” in financial institutions, and why this matters for validation scope
- Regulatory and internal implications: misclassification may lead to under-validation or excessive validation, inefficient resource use
- Reviewing real-world examples: chatbots, AI agents, deterministic methods, ML pipelines and how firms are classifying them
- Developing internal standards and taxonomy: ensuring consistent classification across business lines, jurisdictions, uses
- Evaluating impact on inventory, tiering, validation resources and audit/governance tracking when classification is unclear
14:00 - 14:30
THIRD-PARTY MODEL RISK
Session: Addressing transparency, validation, and oversight challenges in the rise of third-party / vendor AI models
- Setting timely expectations: Third-party Risk Management Frameworks, Vendor obligations and SLAs
- Ability and willingness of Third-party: to provide sufficient documentation and information to reasonably assess adherence to Model Risk Management Framework
- Due diligence best practice: model documentation, data quality management assurance, audit reports, independent validation / review, ongoing performance monitoring
- Integrating into internal frameworks: how to align vendor tool validation with internal model risk framework and regulatory expectations
- Outlining roles of model owners and monitoring teams: how to monitor use and ongoing performance, governance around model changes / version control
14:30 - 15:00
Coffee and networking break
15:00 - 15:40
EXHAUSTIVITY VS. SIMPLICITY
Panel Discussion: What determines the optimal coverage of your portfolio by regulatory models — and where are we heading?
- Assessing the trade-off between model coverage and supervisory expectations as regulators push toward simplified, standardised model landscapes
- Navigating the tension between reducing internal model variability and the rapid expansion of AI-driven modelling techniques across risk, finance, and AML
- Determining when model consolidation improves governance, and when reducing “exhaustivity” creates blind spots, bias, or under-captured risks
- Understanding how the EU AI Act may limit the number of high-risk AI models banks can maintain, and why cross-industry collaboration will be essential to develop efficient, compliant, and scalable model inventories
15:40 - 16:10
PRACTICAL AI FOR GOVERNANCE
Session: Practical AI applications for strengthening model risk management and governance
- Applying AI to enhance model discovery, inventory management, and portfolio-wide risk mapping
- Using NLP and machine learning to automate documentation, audit evidence, and regulatory checks
- Deploying AI-driven monitoring for drift detection, anomaly alerts, performance analytics, and ongoing validation
- Integrating AI-enabled tools into governance frameworks to meet transparency, explainability, and accountability expectations
- Reviewing case studies demonstrating efficiency gains, reduced manual effort, and improved compliance through AI-powered oversight
16:10 - 16:50
TALENT
Panel Discussion: Evolving the MRM function through upskilling, cross-functional talent, and redefining its role in the AI-driven future of risk
- Identifying the skills gap: need for people who combine quantitative modelling, data science/AI, domain risk knowledge and governance acumen
- Building strategies for upskilling existing model risk/validation teams: training programmes, certifications, cross functional rotations
- What collaboration models are used by financial institutions to address skills shortages: enabling first line practitioners, embedding second line oversight, cross pollinating skills across teams?
- Changing role of MRM: from “checking models” to “co creating model risk controls and guiding AI innovation responsibly”.
- Recruitment and retention challenges: how to hire for new competencies while maintaining independence, how to incentivise talent in validation and governance?
16:50 - 17:00
Chair’s closing remarks
17:00
End of day one
Wednesday 18 March 2026
08:00 - 08:50
Registration and breakfast
08:50 - 09:00
Chair’s Opening Remarks
09:00 - 09:35
MODEL RISK SIMULATION
Session: When data driven analysis goes wrong
- Simulation study on estimating macro-economic dependence of credit losses.
- Looking at a dataset with no theoretical correlation, it is investigated how spurious correlation can show up due to p-hacking.
- Gaussian process theory allows to create multiple realistic datasets by controlling the autocorrelation
- Practical suggestions on how expert assessment can combined with data driven methods to reduce spurious correlations
09:35 - 10:10
MACHINE LEARNING
Session: Leveraging machine learning for time series and market risk—balancing innovation with simplicity, speed and robustness
- Applying ML/AI techniques for enhancing time-series analysis in trading book and market risk
- Rediscovering lightweight traditional algorithms for volatility modeling versus deep learning, evaluating trade-offs of speed, interpretability, cost
- Real-world case studies of ML improving back-testing performance, reducing outlier violations, faster model refresh
- Dealing with implementation challenges, such as data noise, regime changes, non-stationarity, model drift, documentation of assumptions
10:10 - 10:45
LIGHTWEIGHT MODELS
Session: Leveraging simple, interpretable models for risk modelling: performance, efficiency and validation trade‑offs
- What are the benefits of simple, interpretable modelling techniques in comparison with deep-learning techniques?
- Case‐studies where legacy algorithms have been “rediscovered” and perform robustly in noisy financial environments
- Defining criteria for algorithm selection: speed, explainability, resource consumption, maintainability, validation burden
- Balancing algorithmic sophistication vs operational practicality considering cost of validation, deployment, oversight and other relevant factors
10:45 - 11:15
Coffee and networking break
11:15 - 11:45
PORTFOLIO RISK MANAGEMENT
Session: Ensemble generalization error as a framework for portfolio risk diversification
- Modelling portfolios as multi-hypothesis structured ensemble models linking each asset to a predictor
- Optimizing a portfolio as a supervised ensemble learning problem, connecting ensemble diversity to risk diversification via information geometry
- Parametric control of out-of-sample portfolio diversification by truncating learning and hypothesis selection, presenting the Quality–Diversity trade-off for return predictions
- The diversity–capacity (complexity) trade-off in ensemble generalization performance and its implications for model selection and parameter tuning for optimal portfolio diversification.
11:45 - 12:15
SCENARIO ANALYSIS AND THE LIMITS OF AI
Session: Designing and validating stress testing frameworks that reflect systemic risks, technical demands, and forward-looking scenarios
- Ensuring models remain valid under extreme conditions and systemic shocks
- Understanding the complexities of interconnected risk modeling, including cascading effects, cross-functional dependencies, and nonlinear stress interactions
- Assessing scenario design through plausibility checks, accurate calibration, relevance to historical data, and adaptability to future conditions
- Evaluating technical capabilities such as revaluation accuracy, computational efficiency, and consistency with front office stress scenarios
- Defining the role of model risk in validating stress testing outcomes and ensuring robust oversight of stress model frameworks
12:15 - 12:55
TECHNOLOGY STRATEGY
Panel Discussion: Bridging risk and front office through scalable infrastructure, real-time modeling, and automation—while upholding robust model risk management
- How model risk teams can ensure alignment between front office pricing engines and risk P&L models without compromising independence or objectivity?
- Identifying the role of cloud infrastructure, scalable computing, and real-time APIs in supporting model monitoring, backtesting, and stress testing in model risk management processes
- What levels of automation are acceptable under MRM standards in automating model inventory, versioning, performance monitoring, and validation workflows?
- Balancing Innovation with Control: winning strategies for maintaining audit trails, “human-in-the-loop” checks, and explainability
- Reviewing real-world cases where risk and front office collaborated on tech upgrades -what worked, what failed, and how MRM ensured model risk didn’t fall through the cracks?
12:55 - 13:55
Networking Lunch
13:55 - 14:25
MULTI-GOVERNANCE SETUP
Session: Enabling AI & MRM teams in one consistent solution
- Establishing a unified governance framework for AI and traditional models
- Creating a single inventory and lifecycle view across AI, ML, and classical models
- Coordinating workflows, approvals, and accountability across AI and MRM teams
- Aligning AI principles and MRM controls within one governance solution
- Embedding governance activities into development, validation, and monitoring processes
- Demonstrating efficiency and consistency gains from a multi-governance approach
14:25 - 14:55
GENAI IN CREDIT RISK & AUTOMATION
Session: Model risk perspectives on scaling GenAI in credit processes: validation, traceability, and regulatory compliance
- Evaluating GenAI models used in credit risk applications and the role of model risk teams in KYC, credit scoring, document analysis, and customer interactions
- Examining roll-out strategies from a model risk standpoint, including timing of MRM involvement, embedding controls early, and aligning validation with development cycles
- Managing modular and bespoke GenAI tools while considering their impact on validation scope, documentation, and integration into existing systems
- Addressing regulatory classification of high-risk GenAI use cases and how model risk teams are adjusting frameworks to meet new external and internal compliance demands
14:55 - 15:25
Coffee and networking break
15:25 - 15:55
VALIDATION
Session: Building adaptive validation frameworks for dynamic AI and agentic models; Addressing complexity, risk tiering, and continuous monitoring
- Designing agile and modular validation frameworks suited to rapidly evolving AI/ML and agentic systems (rather than static models)
- Embedding validation into development lifecycle; ongoing testing, monitoring, drift detection, post-deployment validation loops
- Risk-tiering of AI/ML models to differentiate high-risk vs low-risk and the effect on validation
- How to deal with coordination between agents, time dependencies, cascading decision chains, emergent behavior and other issues specific to agentic AI?
15:55 - 16:35
COLLABORATION
Panel Discussion: Exploring how model risk teams can evolve through shared learning, research collaboration, and systematic benchmarking
- How banks are building internal research capacity to explore AI, data, and risk model innovation?
- Opportunities to benchmark models across institutions without breaching confidentiality
- Best practices of how to make learning continuous through internal training, model risk certifications, and talent refresh strategies
- How to capture and circulate insights from success and failure by embedding knowledge management into the MRM function?
16:35 - 16:45
Chair’s closing remarks
16:45