On 30 March 2026, the banking supervision arm of the European Central Bank (ECB) published a decision that disappears under the catch-all label of cutting red tape in the public eye, yet is causing movement in banks' risk departments: from 1 October 2026, institutions may implement material changes to their internal credit risk models as soon as they have submitted a complete application – without waiting for the supervisor's prior approval, previously mandatory. The Internal Ratings-Based Approach (IRB), with which large European banks calculate their own risk weights, is not being relaxed, but its supervision is being turned upside down. A system in which approval was the precondition for implementation becomes a system in which control of the implementation is downstream. The step is accompanied by revised materiality criteria from the European Banking Authority (EBA), published on the same day. Anyone seeking to gauge the significance as a risk or modelling officer should keep two things apart: what changes operationally, and what price the acceleration carries.
What: From 1 October 2026, for material changes to internal credit risk models the ECB replaces prior approval (ex ante) with downstream control (ex post); announced on 30 March 2026
Condition: The bank's internal control function – validation, internal audit and model risk together – must credibly confirm the revised model's compliance with regulatory requirements before it is implemented
Floors: If a change lowers risk weights, a floor of 98 per cent applies to material model changes (maximum 2 per cent capital relief) and of 100 per cent to material model extensions (no relief); the floor is lifted only after a targeted on-site investigation
EBA side: The revised Regulatory Technical Standards (RTS) rely more heavily on quantitative thresholds and limit qualitative triggers; they exist as a final draft but are not yet in the EU Official Journal and therefore not formally in force
Context: Part of the ECB banking supervision's "Next Level Supervision" project; the basis of trust is a decade of model hygiene since the Targeted Review of Internal Models (TRIM)
What Changes on 1 October
Until now, every material model change followed the same path: the bank submitted an application, the ECB examined it, and only after approval could the revised model be used in the capital calculation. In practice, according to an estimate by the consultancy Advisori, this meant waiting times of twelve to eighteen months – there is no official ECB average for this; the supervisor itself speaks only qualitatively of procedures that take too long. During that time, the institution had to keep running the old model even though it already considered the new one superior. It is precisely this bottleneck that the reform addresses.
In future, the following applies to material changes: the bank submits a complete application package and may implement the revised model shortly afterwards, provided its internal control function credibly confirms compliance with regulatory requirements. What the ECB understands in detail by a complete application package is not conclusively defined in the accessible primary documents – this is likely to become one of the first operational clarification tasks for institutions. The decisive shift is in the sequence: the supervisor no longer examines before implementation, but after it. Accordingly, material model changes no longer automatically trigger an on-site investigation. For scale: in 2025 the ECB conducted 74 internal model investigations, more than 90 per cent of them prompted by banks' applications for initial approval or material change. In future, the supervisor decides on a risk basis where to look more closely.
The Two Thresholds: 98 and 100 Per Cent
The sensitive point of any ex-post logic is the phase between implementation and examination. As long as the ECB has not yet investigated a revised model on site, it should not provide any unexamined capital relief. For this, the reform introduces two floors that are often condensed into a single one in the debate, but in fact differ in height. If a material model change lowers risk-weighted assets (RWA), a floor of 98 per cent applies: the bank may book at most two per cent of the calculated relief immediately. For a material model extension – the expansion of a model to new portfolios – the floor is 100 per cent, so there is initially no relief at all.
In both cases, the floor is lifted only once the ECB has thoroughly assessed the features of the new model in a targeted on-site investigation. This keeps the acceleration real for the bank – it can deploy the better model immediately – but the capital effect takes full hold only after supervisory confirmation. Operationally, this means that the ability to calculate both figures in parallel, the model result and the floored value, remains mandatory for institutions, even though the parallel runs of the old approval logic fall away.
The Reform's Second Half: the EBA Criteria
The ECB governs the procedure, the EBA the definition. On the same 30 March 2026, the European Banking Authority published its final report with revised Regulatory Technical Standards on the question of when a model change counts as material at all. The legal basis is Article 143(5) of the Capital Requirements Regulation (CRR) as amended by CRR III; the new standards amend Delegated Regulation (EU) No 529/2014. The thrust: a stronger reliance on quantitative thresholds and a marked reduction of qualitative triggers. Qualitative triggers are to be confined in future to what actually rebuilds a model – changes of methodology, recalibrations of risk parameters or changes to the definition of default. Routine maintenance, by contrast, will generally be subject only to a notification requirement, provided it does not breach the quantitative thresholds.
In sum, this shifts a considerable part of what banks previously had to treat as material and therefore subject to approval into the realm of mere notification. An important caveat belongs here, however: the EBA standards are a final draft. Before they become legally binding, the EU Commission must endorse them and publish them in the Official Journal of the European Union, which experience shows takes several months. It is therefore not ruled out that the new ECB process starts on 1 October 2026 while the formal applicability of the revised materiality criteria is still pending – a timing gap that institutions should factor into their planning.
Why the ECB Is Letting Go Now
The step does not come out of nowhere. It stands at the end of a decade in which the supervisor first scrutinised banks' internal models systematically. The Targeted Review of Internal Models, TRIM for short, ran from 2016 to 2021 and comprised some 200 on-site investigations at 65 significant institutions (SIs). From the ECB's perspective, this effort has improved the models' compliance and the quality of the banks' internal control functions to such a degree that it now feels able to supervise in a more targeted, risk-based way. In other words: the supervisor is letting go because it now trusts the model hygiene it enforced itself.
The reform is embedded in the "Next Level Supervision" project, which the ECB launched in 2025 and is implementing in 2026. It applies the simplification objective to all major supervisory tasks – authorisations, model approvals, stress testing, reporting and on-site inspections. Claudia Buch, Chair of the ECB's Supervisory Board, had already outlined the logic in mid-2025: good internal models are key to good risk management, but maintaining and validating them consumes time and resources at banks and supervisors alike. Streamlining the model landscape can therefore contribute considerably to simplification without making the banking sector less resilient. The three-tier classification – material, non-material with prior notification, non-material with subsequent notification – remains structurally intact. What changes is not the framework, but the question of what triggers materiality.
The Counter-Argument
An honest assessment also names what the gain in convenience costs. Three points deserve attention. First, the timing: the simplification agenda falls in a phase of heightened macro-financial uncertainty – geopolitical tensions, an uncertain rate path, latent recession risks. Observers such as Finance Watch and academic voices have noted that supervision is thus being loosened precisely when risks are rather rising. That is no contradiction to the reform, but a context that raises expectations of the downstream control.
Second, and more weighty, the relocation of the first quality control: it moves from the supervisory authority to the bank's own control function. Whoever wants to use the fast track needs their validation, internal audit and model risk to confirm compliance before the ECB looks. This sharpens a conflict of interest that previously rested with the supervisor: the function meant to confirm a more favourable risk weight sits in the same house as the business units that benefit from that risk weight. This is precisely where the core risk of the ex-post approach lies – and precisely where it is decided whether a bank can realise the speed advantage at all.
Third, the pro-cyclicality of the floors. The 98 per cent floor falls only after a completed on-site investigation. If many institutions submit risk-weight-lowering model changes at the same time – for instance in a downturn, when capital relief is particularly attractive – and the ECB cannot scale its examination capacity accordingly, the floor stays active longer than intended. The hoped-for capital effect then shifts back precisely at the moment it would be needed most. Added to this is a grey area: the ECB reserves the right to stay with the classic ex-ante procedure for "sensitive cases". What counts as sensitive is not conclusively defined. For banks, this means planning uncertainty about whether an application lands in the fast or the standard track.
What Risk and Modelling Teams Should Do Now
A good three months remain until 1 October 2026. Four work packages determine whether an institution actually captures the reform's benefit or merely carries the uncertainty.
By October 2026: Many historical materiality decisions will come out differently under the revised, more quantitative thresholds. Institutions should review their inventory of model changes to determine which will in future be merely notifiable and which remain in the approval-adjacent zone. Doing the reclassification early avoids misclassifications that later return as a supervisory finding.
By October 2026: The fast track stands or falls with the credible confirmation by validation, internal audit and model risk. That requires documented, mutually independent review processes and a clear separation from the business units that benefit from the model result. A weakly positioned control function loses the reform's main advantage – it effectively forces the bank back into the slower procedure.
By Q4 2026: Even without the old parallel runs, every institution must be able to calculate the 98 or 100 per cent floor cleanly and reflect it in the capital report – including the logic of when the floor falls away after an on-site investigation. This capability belongs in the model and reporting infrastructure, not in a manual side calculation.
Ongoing: As long as "sensitive cases" are undefined, institutions should align early with their Joint Supervisory Team on which planned changes are likely to land in the fast and which in the classic procedure. This certainty of expectation is relevant for planning: it determines whether a capital effect lies within the quarterly horizon or beyond a year.
Risks and Open Questions
Three caveats belong to a sober evaluation. First, applicability: the ECB process is finally adopted as supervisory administrative practice and binding from 1 October 2026, whereas the accompanying EBA standards are still awaiting endorsement by the EU Commission – a time lag that can diverge in implementation. Second, the data basis: the often-cited twelve to eighteen months of current processing time stem from a consultancy assessment, not from an official ECB statistic; nor can the number of institutions currently using IRB models be precisely quantified from the public documents. Third, the shift in incentives: by moving the first compliance check from the supervisor to the bank, the resilience of the entire system hangs on the independence of the internal control function – and thus on a body that structurally sits closer to the business than the supervisor.
The strategic consequence: the reform is no gift to the banks, but a shift of responsibility. The ECB returns speed and in exchange holds the internal control function more firmly to account. Institutions that set up their validation, their floor logic and their materiality classification cleanly now will make model changes effective in months rather than years. The others carry the same speed expectation from the supervisor without being able to cash in the advantage.
Christian Schablitzki
Strategy & Management Consultant · Agentic AI expert for financial institutions
Over 20 years in investment banking and derivatives trading, followed by more than 10 years advising financial institutions. Currently a Partner at Infosys Consulting in Germany. Certified in Google AI, Generative AI Leader (Google Cloud) and IBM RAG and Agentic AI.
LinkedIn profile →Read on – every 14 days in your inbox.
Capital markets insights, regulatory updates and AI trends. Concise, well-founded, free.
GDPR-compliant. Unsubscribe at any time.