Neil Catton
A joint article with Lukas Ziewer, Senior Advisor, Risk Strategists, Regulatory Solutions

When the model said it was fine.

Three decades of model risk governance

1996Basel’s Market Risk Amendment sets the foundations: board accountability, backtesting, stress testing.2011SR 11-7 becomes the first comprehensive model risk framework.2026Its revision, SR 26-2, explicitly excludes generative and agentic AI from scope.

The human moment

Every framework lands, in the end, on one signature.

  1. We know what models we are running
  2. They behave as validated
  3. Validation is current in substance, not only in date
  4. No model decides without a human review point
  5. The expertise to challenge remains in-house

That is what an attestation of effective model risk management implicitly claims. Current deployment conditions give reasonable cause to doubt all five.

Why it matters now

Exclusion from scope is not exclusion from consequence.

The models are deployed. The decisions are being made.

USParked. Out of scope, a Request for Information to follow, no timeline set.
UKAlready inside the perimeter, under principles written before these systems existed.
EUAnswered by restating the frameworks supervisors already have.
The argument

The attestation can be sincere and wrong.

The signature attests to the machinery.

The machinery no longer reaches the risk.

If your name is on the attestation

Ask what your signature now claims.

  1. ReadTwenty-one pages, for a reader who understands financial services governance and wants the AI-era gap named plainly.Download the PDF
  2. TalkOne conversation about your own attestation, and whether the question applies to it.Book a meeting

Co-authored for joint publication rather than commissioned as a client briefing. The EIOPA Opinion, the ECB’s February 2026 keynote, the scope of PRA SS1/23, the Durham HART validation study and the ENTSO-E final report on the Iberian blackout were each read in full against the primary publication during review. Published 7 August 2026, revised 17 August 2026.

If you want the detail

What’s in the briefing

What’s inside the twenty-one pages

For SMF holders, Chief Risk Officers and board risk committee members who carry personal accountability for model risk, and practitioners in healthcare, criminal justice, energy and the public sector now meeting the same problems:

  • What Model Risk Management was built to do: the regulatory genealogy from the 1996 Basel Market Risk Amendment through SR 11-7, Solvency II and the conduct-of-business rulings
  • Why financial services developed the discipline before any other sector: leverage, model centrality, and systemic risk with an implicit state guarantee
  • Five specific ways AI changes the conditions the frameworks assume: speed and “vibe coding”, self-modification, direct-to-execution, documentation decay, and the classification problem
  • Three regulatory postures compared: the US’s explicit exclusion under SR 26-2, the UK’s technology-agnostic incompleteness under SS1/23, and Europe’s reabsorption of AI into existing requirements
  • What the same failure modes look like in healthcare, criminal justice, energy and utilities, and public sector algorithms, and what each side can learn from the other
  • The accountability question: what an attestation implicitly claims, and why the named individuals who sign it have reasonable cause for doubt
  • A five-point work programme for the model risk management profession to examine next

Questions people ask before reading

Is this vendor material?

No. It is an independent, jointly-authored analysis. Regulatory positions were each read in full against the primary publication, not taken from summaries. Foundational legal and regulatory sources are drawn from the established legal record; secondary industry analyses are cited for context only, not as primary evidence for any claim.

Who is Lukas Ziewer?

An independent risk strategist with over twenty-five years of leadership in financial risk management and actuarial work across EMEA and Bermuda, including Chief Risk Officer and Partner roles at Athora Group, MetLife, KPMG and Oliver Wyman. Now operates through Fuseki Risk Insights.

Is it free to download?

Yes. It downloads directly, with no form and no email address required.

How current is the research?

Published 7 August 2026 and revised 17 August 2026. Regulatory positions described are current as at revision and may have moved again since, so verify against the original source before relying on any figure.

Can I share it with my board or team?

Yes. It is written for exactly that circulation. The document itself asks that it not be redistributed externally beyond that without permission from the authors.