Book a demo
Topics · REGULATORY

Asset Management Model Risk Quarterly Review | Moring

A quarterly model risk review in asset management is the recurring evidence-based examination of every deployed model and agent: drift results, incident and exception history, tier changes, access grants, and vendor model behavior. Research supports moving from periodic validation to continuous governance, with the quarterly cycle aggregating that continuous evidence for owners, risk committees, and boards.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading
  5. Sources

Short answer

A quarterly model risk review in asset management is the recurring evidence-based examination of every deployed model and agent: drift results, incident and exception history, tier changes, access grants, and vendor model behavior. Research supports moving from periodic validation to continuous governance, with the quarterly cycle aggregating that continuous evidence for owners, risk committees, and boards.

How it works

The quarterly review has long been the ritual of model risk, but its content is changing. Under a periodic-validation mindset, an asset manager refreshes documentation and reruns backtests. Under the continuous-governance framing that industry white papers now advocate, the quarterly review aggregates living evidence: drift surveillance results, agent action logs, incident and exception records, access changes, and version histories accumulated since the last cycle.

What belongs on the agenda is concrete. AI-MRM research describes the discipline as identifying, measuring, monitoring, and controlling risk across the model lifecycle, extending SR 11-7 principles to generative and agentic AI through drift detection, model tiering, and attestation. A quarterly cycle for an asset manager should therefore confirm that tiering still matches decision impact, that monitoring caught what it should have, and that attestations are current.

Agent activity deserves its own section. Because chained agents compound risk by taking actions without human review, and governance programs often fail at enforcement with agents shipping without guardrails, the quarterly review should examine runtime evidence: what agents accessed, which calls exceeded limits, where approvals were bypassed, and whether agent authority definitions still reflect the firm's investment process.

Vendor and third-party models need explicit attention. Model risk in the agentic era includes models that change behavior behind stable APIs, and lifecycle research warns that a model performing well in testing can degrade silently in production with nobody noticing until damage is done. Quarterly comparison of output distributions, cost patterns, and version changes for externally hosted models gives an asset manager an early view of that decay.

The regulatory backdrop raises the stakes of the cycle. SR 26-2, adopted April 17, 2026 by the Federal Reserve, the OCC, and the FDIC, replaced SR 11-7 in the first model risk guidance overhaul in fifteen years, and commentary notes it regulates models rather than agents, so agent review content is the firm's design choice. EU AI Act obligations add periodic-review expectations for firms with European operations.

A review that produces only a narrative misses the point; the output should be decisions. Reclassify models whose drift patterns changed, tighten permissions where agent behavior surprised reviewers, retire models that no longer earn their monitoring cost, and record every exception with an owner and date. Governance research stresses mechanisms and evidence over documents, and the quarterly cycle is where an asset manager converts accumulated evidence into visible control.

Frequently asked questions

What should a quarterly model risk review cover for agents?

Beyond model drift and performance, review agent runtime evidence: systems accessed, calls exceeding limits, approval bypasses, and whether agent authority definitions still match the investment process. Because chained agents act without human review, quarterly agent review is where asset managers catch behavior that per-model monitoring misses.

How does continuous governance change the quarterly review?

Continuous governance means the quarterly cycle aggregates evidence rather than manufactures it: drift surveillance results, action logs, access changes, and version histories gathered all quarter. White papers on model risk in the AI era describe exactly this shift from periodic validation to continuous governance, with the quarterly meeting as the decision point.

What should the review decide, not just document?

Concrete decisions: reclassify tiers where impact changed, tighten agent permissions where behavior surprised reviewers, retire models that no longer justify monitoring cost, and register exceptions with owners and dates. Governance research emphasizes that controls, not policy documents, reduce model risk, so each cycle should leave the control set changed where evidence demands it.

Why include vendor models in the quarterly cycle?

Externally hosted models can change behavior behind stable APIs, and lifecycle research warns that production models can degrade silently until damage surfaces. Quarterly comparison of output distributions, costs, and version changes gives an asset manager early detection of vendor-side drift that internal testing would miss.

Sources

  1. - federalreserve.gov
  2. - federalreserve.gov
  3. - artificialintelligenceact.eu