Risk April 2026 7 min read

Model Risk
Management:
A Global Bank's
Two-Regime Problem

On 17 April 2026 the US agencies rescinded SR 11-7 and replaced it with a principles-based, explicitly non-enforceable framework. In London, SS1/23 continues to bind UK banks to a much firmer standard. For globally active institutions, the consolidated MRM framework now has to span a widening gap.

By Georgios Petropoulos  |  LINXS Advisory
2026
Year SR 11-7 rescinded
5
PRA SS1/23 principles
$30B
Asset threshold under new US guidance

01 Two Regimes, One Bank

Until April 2026, SR 11-7 and SS1/23 were broadly compatible — both principles-based, both organised around development, validation, governance and use. The new US interagency guidance breaks that symmetry. For groups regulated on both sides of the Atlantic, the practical reality is now two materially different supervisory philosophies sitting over a single inventory of models.

United States — April 2026
OCC, Federal Reserve and FDIC have replaced SR 11-7 with a concise, principles-based statement. Non-compliance will not result in supervisory criticism. Primary applicability is to firms above $30B in assets.
United Kingdom — SS1/23
PRA's five principles have been in force since May 2024. Model risk is treated as a risk in its own right, with a designated senior manager (SMF) accountable at board level and mandatory annual self-assessments.
AI & Generative Models
The new US guidance explicitly excludes generative and agentic AI, with an RFI to follow. SS1/23 makes no such carve-out — AI/ML sits within the existing model definition and is supervised accordingly.
Validation Independence
The US text now states that validation quality depends on rigour, not org structure. The PRA continues to expect structural separation, with independent challenge of developers and users embedded in the second line.
Scope of "Model"
US definition has narrowed in practical effect. SS1/23 reaches further: financial reporting models, vendor models, and material deterministic quantitative methods all sit inside the perimeter.
Supervisory Posture
US guidance is non-enforceable on its face. UK firms face self-assessments, s166 reviews and active dialogue with the PRA on remediation plans running over multi-year horizons.

02 Where the Friction Lives

A global bank cannot operate two MRM frameworks in parallel — the duplication, governance overhead and audit cost are prohibitive. In practice, the group framework calibrates to the higher bar. SS1/23 is now that bar. The challenges below are the points where calibration meets resistance.

1
Scope Mismatch
Inventories built for SR 11-7 are too narrow for SS1/23.
Identifying and tiering deterministic quantitative methods, EUCs and financial-reporting models is a multi-year undertaking. A US inventory built around the new guidance gives the group no relief — the UK perimeter must still be populated.
2
The AI / GenAI Void
In scope in London, parked in Washington.
The US carve-out for generative and agentic AI cannot be mirrored in a group policy without breaching SS1/23. Firms either apply UK standards globally to AI/ML, or carry an awkward exception register for US-only deployments.
3
Validation Independence
Structural separation is no longer a US given.
Where the new US guidance softens reporting-line and compensation independence, the PRA does not. Group standards must preserve structural independence even where the local US regulator would accept less.
4
Board Accountability
SMF designation has no US analogue.
UK boards are personally on the hook through the SMF regime. The new US text steps back from prescriptive board duties. Group committee charters and MI packs must continue to satisfy the sharper UK standard.
5
Evidencing the Framework
Self-assessment is a UK obligation, not a US one.
Annual SS1/23 self-assessments, remediation plans and s166 readiness require continuous documentation. US examiners will not ask for the same artefacts — but the evidence base has to be produced for the group regardless.

03 Side by Side

A practitioner-level view of where the two regimes converge and where they pull apart. Group MRM functions should treat the right column as the operating standard.

US — Apr 2026 UK — PRA SS1/23
Style Concise, principles-based, non-enforceable Principles-based but actively supervised
Scope of "model" Narrower; AI/GenAI excluded pending RFI Broad; FR models, DQMs, AI/ML included
Applicability >$30B firms; smaller banks where material Firms with internal model approval; expected to expand
Validation Quality depends on rigour, not structure Structural independence retained
Board / governance High-level principles; no enumerated duties SMF accountability; annual self-assessment
Enforcement Non-compliance not a basis for criticism s166, remediation plans, ongoing dialogue
Practitioner Insight

For globally active banks, the April 2026 US update is not a relief valve. It is a divergence.

SS1/23 has effectively become the binding constraint on group MRM frameworks.

The right operating answer is to hold the UK standard globally and document local US accommodations as managed exceptions, not as the baseline.

Recalibrating your group MRM framework?

LINXS Advisory works with banks on regulatory gap analysis, SS1/23 self-assessments, and group-wide model risk uplift across the US, UK and EU.

Get in Touch Visit LINXS Advisory