Discover the impact of future AI Regulation with Yu Pan, Chief Model Risk Officer at US Bank, from our panel session ELC Annual 2023 - The largest conference for engineering leaders in San Francisco!
This is the third article in our 3-part series where we uncover the insights from our panel discussion at ELC Annual 2023.
You can read the first articles with Grace Chu, former Senior Product Counsel at Adobe, on Responsible AI Development: Working Hand-In-Hand With Legal and Sushant Hiray, Sr. Director of ML at RingCentral, on How to Future-Proof Your AI Projects and Implement Governance
In this blog post, we will uncover:
- Navigating AI Compliance in Banking
- Expert Perspectives on Model Risk Management
- Tech Industry Advice for Banking Compliance
Let’s dive in!
About Yu Pan
As the Chief Model Risk Officer, Yu is responsible for developing and leading the enterprise model risk functions, identifying and managing model risk, and overseeing the assessment/validation of all the models/tools used across the bank.
Developing and deploying AI models in the banking industry comes with substantial compliance requirements and challenges, as discussed by Yu Pan, an expert in model risk management at US Bank. Here are key insights from Pan on navigating model development and documentation processes.
What processes do you have in place to ensure AI model compliance?
In discussing the model development process in banking, Yu Pan provides a comprehensive overview of the multiple steps models must go through. These include conceptual design, data processing, testing, validation, implementation, monitoring, annual reviews, and revalidation every 2-3 years.
"Whatever models we build that impact customers are not only subject to the highest validation security but also heavily examined by the regulators as well."
To meet requirements, banks must thoroughly demonstrate that the model serves its intended purpose, explore alternative approaches, validate the clean and unbiased data, and rigorously test model performance. The extensive process requires significant investments in model risk management teams to ensure that the models are conceptually sound, perform robustly, and meet compliance requirements.
When discussing models and systems with regulators, what is US Bank's documentation scale?
Pan emphasizes how critical thorough documentation is for banking models.
"Documentation is very critical, as we go through the model lifecycle you must demonstrate your understanding of the business and the quality of your model."
He notes that documentation can easily exceed hundreds of pages to meet model risk management standards and regulator expectations for one high-risk model alone. Banks must clearly explain models and justify development choices to regulators.
What is the most important takeaway on this topic for the team and audience here today?
Offering guidance to tech companies aiming to provide services for banks, Pan advises
"If you want to work with the heavily regulated banking industry, you need to understand the risk management requirements and regulatory rules, and follow them. That is my advice."
While tech firms develop leading-edge tools and models, Pan stresses they must closely adhere to banking regulations for their products to be usable in this industry. AI model innovation must fit within banking compliance frameworks.
Yu Pan’s insights on documenting model development in the banking industry:
- Banks must follow a rigorous process for AI model compliance.
- Extensive documentation is crucial for transparency and regulatory compliance.
- Tech firms must understand and adhere to banking regulations for success in the industry.
Overall, Pan emphasized that while AI will drive the future of banking, building and implementing AI models in the heavily regulated banking industry involves comprehensive processes, extensive testing and documentation, and strict compliance requirements. Tech companies must closely follow the risk and compliance rules if they want their models and services adopted.