Asset management AI model risk management SR 26-2 | Moring
Senior compliance and technology leaders in asset management navigating ai model risk management must evaluate how Supervisory Letter SR 26-2 alters model risk frameworks. While SR 26-2 excludes generative and agentic artificial intelligence from its scope as novel and rapidly evolving, asset management firms must still apply existing risk management practices and rigorous governance to address complex AI model risk management challenges.
Short answer
Senior compliance and technology leaders in asset management navigating ai model risk management must evaluate how Supervisory Letter SR 26-2 alters model risk frameworks. While SR 26-2 excludes generative and agentic artificial intelligence from its scope as novel and rapidly evolving, asset management firms must still apply existing risk management practices and rigorous governance to address complex AI model risk management challenges.
How it works
For senior compliance and technology leaders in asset management, understanding ai model risk management under Supervisory Letter SR 26-2 is critical as modeling complexity grows. Asset management institutions leverage innovative technological approaches to improve efficiencies and better address specific risks, making model risk management frameworks central to daily operations and significant business lines.
SR 26-2 supersedes SR 11-7 and establishes principles for effective model risk management while recognizing that practices vary based on risk profiles and model usage. Within asset management, ai model risk management requires structuring governance practices that align with an organization's size, nature, and complexity, ensuring that models supporting significant business lines receive appropriate oversight without misinterpreting the boundaries of regulatory guidance.
Federal Reserve, OCC, and FDIC guidance emphasizes that SR 26-2 does not set forth enforceable standards or prescriptive requirements, meaning non-compliance does not automatically result in supervisory criticism. However, Deloitte, KPMG, EY, and Moody's note that deficiencies in ai model risk management can still draw supervisory attention for unsafe or unsound practices. Furthermore, while generative AI and agentic AI models are excluded from SR 26-2 due to being novel and rapidly evolving, firms must still use existing risk management and governance practices to oversee these systems.
Practitioners in asset management building an ai model risk management program should implement risk-based tiering, utilize the immaterial designation for expedited oversight of low-risk models, and narrow inventory definitions to exclude simple arithmetic. Leaders must also account for model risk management principles across the model lifecycle, ensuring independent validation, ongoing monitoring, and clear documentation are maintained across all relevant asset management systems.
Failure modes in asset management ai model risk management include relying on static validation cycles for complex systems, failing to monitor third-party infrastructure changes, and lacking kill-switch protocols or robust incident response architectures. As noted in industry surveys, many banking and financial professionals lack confidence in their ability to shut down malfunctioning AI models or report failures, highlighting vulnerabilities that asset management firms must actively mitigate to prevent regulatory and operational harm.
Asset management teams should begin their ai model risk management evolution by auditing their current model inventory against SR 26-2 definitions, establishing risk-based tiering to prioritize resources, and implementing continuous monitoring guardrails for advanced technologies. By aligning internal risk governance with supervisory expectations, asset management institutions can build value-additive model risk management functions that reinforce stakeholder trust while managing model velocity.
Frequently asked questions
How does SR 26-2 impact ai model risk management within asset management firms?
Supervisory Letter SR 26-2 impacts ai model risk management in asset management by formalizing risk-based tiering and introducing an immaterial designation for expedited oversight. While generative and agentic artificial intelligence models are excluded from SR 26-2 due to being novel and rapidly evolving, asset management organizations are still expected to apply existing risk management and governance practices to maintain rigorous control over their technology portfolios.
Are generative AI systems covered under SR 26-2 for asset management ai model risk management?
No, generative AI and agentic AI models are explicitly excluded from the scope of SR 26-2 because they are novel and rapidly evolving. However, senior compliance and technology leaders in asset management must recognize that this exclusion is a regulatory boundary rather than a governance exemption, requiring robust ai model risk management practices aligned with broader supervisory expectations.
What are the primary operational challenges for asset management ai model risk management under the new guidance?
Key operational challenges for asset management ai model risk management include addressing the opacity and stochasticity of advanced systems, managing third-party infrastructure dependencies, and overcoming a lack of confidence in kill-switch protocols. Asset management firms must move beyond static validation to implement continuous governance and behavioral testing for complex analytical tools.
How should asset management leaders structure their ai model risk management teams in response to SR 26-2?
Asset management leaders should structure their ai model risk management functions by leveraging SR 26-2 to de-scope deterministic calculations, establishing risk-based tiering for efficient resource allocation, and ensuring independent validation. This transformation allows asset management firms to build leaner, risk-driven oversight mechanisms that enable model velocity while maintaining regulatory alignment.