8 Lessons in Model Risk Management from JP Morgan Chase CEO Jamie Dimon

In a recent Bloomberg interview, JP Morgan Chase CEO Jamie Dimon shared his perspective on the current state of global financial markets. This included his views on artificial intelligence, machine learning, large language models, and risk management.

From his talk, which can be found here, we pulled eight lessons that model risk management professionals should consider:

Be Prepared … for Everything: Dimon emphasizes the critical nature of preparedness prior to a crisis occurring, rather than responding reactively. This preparedness extends to managing risks associated with machine learning models. Whether focusing on data integrity and bias mitigation or model transparency and explainability, organizations can better prepare to manage risks associated with ML models, leading to safer and more effective AI deployments.

Understand the Broad Range of Potential Outcomes: Effective risk management in machine learning involves anticipating a wide range of potential outcomes. Planning should include scenarios that could have significant negative impacts. “You can’t be prepared after the fact,” Dimon tells Bloomberg Originals host and executive producer Emily Chang. “You’ve got to be prepared before the fact.”

Crisis Management Should Be Considered for Models, Too: The JP Morgan Chase philosophy of having a “fortress balance sheet” demonstrates the approach of maintaining strong capital and liquidity to withstand crises. This principle can be adapted for machine learning models, ensuring they have robust fail-safes and recovery mechanisms. “When I talk about risk management, it’s always about the range of outcomes,” he tells Bloomberg.

Prepare for Regular Updates and Monitoring: Dimon stresses the need to stay informed and updated on global affairs and business practices. This parallels the need for continuous monitoring and updating of machine learning models to stay relevant and effective. Machine learning model accuracy can decrease over time as the data they were trained on becomes less representative of the current environment (model drift). Plans should consider regular model re-evaluation and updating procedures, with the ability to roll back to previous versions if necessary.

Consider the Impact of Geopolitical and Economic Changes: Understanding the broader geopolitical and economic context can influence machine learning strategies, as these factors can affect data integrity and model outputs. MRM teams should plan to maintain thorough documentation of data sources, model decisions, and the rationale behind them. Regular audits can ensure ongoing compliance and help identify any influences that changing geopolitical or economic factors may have on your models.

Anticipate Unexpected Events: Planning for machine learning involves considering unlikely but possible events, such as cyberattacks or sudden regulatory changes, which could impact model functionality. Consider a wide range of potential events, including low-probability but high-impact scenarios, and ensure that your plans to address them stay current.

Regulatory Considerations and Guardrails: Dimon stresses the importance of implementing regulatory considerations and guardrails, especially as technology evolves. This is crucial in machine learning to prevent misuse and ensure the productive and safe use of technology. “You want to have plenty of innovation,” Dimon tells Bloomberg, “(but) you need guardrails.”

Continuously Integrate New Technologies Safely, Thoughtfully: Dimon highlights some of the ways AI and machine learning can integrate into AI and machine learning in banking processes. Technology can improve efficiency and decision-making, he says, which also applies to how machine learning models should be integrated thoughtfully into existing systems to enhance performance without compromising stability. 

Ready to drive best practices throughout your model risk management lifecycle? ValidMind can help. Book a demo today to learn how your teams can boost efficiency, speed model time to market, and help your MRM teams reach their full potential.