The duty already applies: What the EBA reporting framework for ISDA SIMM sets out

On 5 August 2026 the European Banking Authority (EBA) published Consultation Paper EBA/CP/2026/16. It sets out which data counterparties will have to report to supervisors on their use of the Standard Initial Margin Model (SIMM) of the International Swaps and Derivatives Association (ISDA). The market uses this model to calculate initial margin, the collateral posted against non-centrally cleared over-the-counter (OTC) derivatives. According to the draft, it applies to every financial and non-financial counterparty that runs a SIMM-based model under the European Market Infrastructure Regulation (EMIR). Comments are due by 2 November, a week after this article appears.

The EBA's press release leads with two messages: a new reporting framework, and substantially lighter requirements for firms with less significant OTC activity, which would report only once a year. Read it as a new obligation with a discount for size and you miss half the design. Firms already have to supply the data. Since 1 March 2026, Article 10 of EBA Decision EBA/DC/610 has required every user of the model to provide the EBA with information on its use, performance, calibration, disputes and add-ons. The August draft settles how often, in what format and for whom that obligation becomes routine. And the discount depends on the derivatives volume of the group, whatever licence the individual entity holds.

The duty to supply data has applied since 1 March

EMIR 3, Regulation (EU) 2024/2987, set the framework and has applied since 24 December 2024. Under Article 11(3), counterparties must for the first time obtain approval from their competent authority before using or changing an initial margin model. Where the model is based on a so-called pro forma model, it also needs validation by the EBA, because the regulation allows authorities to approve it only “where the pro forma model has been validated by EBA”. According to the EBA, exactly one such industry model exists: SIMM, in use since September 2016. Article 11(12a) created a central validation function at the EBA. The authority announced its launch on 26 February 2026, and Decision EBA/DC/610 entered into force on 1 March. On that day the EBA validated nothing yet and opened its system instead. In the same press release it said it expected to issue its first validation decision, together with the list of counterparties covered, in the fourth quarter of 2026. As of 18 October, when research for this article closed, no such decision had appeared.

The February decision already describes the data in some detail. Article 10(5) of EBA/DC/610 covers information on the use, performance and calibration of the model, on margins calculated under the standardised grid and under models, on explicit choices not to use a model, statistics on reconciliations and disputes, details of the most important disputes, and data on add-ons. What is missing is structure. The consultation paper puts it plainly: “Currently, this information is not collected or reported in a regular, structured and harmonised manner across reporting entities.” The framework is meant to change that, and it does so in a strikingly lean legal form. The outcome will be an EBA decision “that will be directly applicable”, as the paper states. The draft’s signature line is reserved for the EBA’s Executive Director. Its basis is Article 11(3)(b) of Decision EBA/DC/610, which empowers the Executive Director to set the scope, format, frequency and reference dates of the data.

Formally, this is an implementing measure that never passes through the European Commission or the European Parliament, which puts the weight of the process on the consultation. It is the only formal point at which scope and frequency are negotiated in public. Afterwards, the EBA has said, it will calibrate the open threshold using the 2026 fee data, and the technical specification will follow later. Anyone with something to say on the paper's twelve questions has to say it now, or wait for the next revision.

The EBA monitors what it does not revalidate

The framework only makes sense when set against the model's own rhythm. Since 2025, ISDA has recalibrated SIMM twice a year. Scott O'Malia, Chief Executive Officer of ISDA, announced on 7 September 2023 that from 2025 the model would be calibrated “on a semiannual basis”, “in coordination with global policymakers”. According to ISDA, the current version, 2.8+2512, has applied since 11 July 2026. For the EBA, an update of this kind will usually not trigger a validation. Article 3(6) of Decision EBA/DC/610 makes clear that updates to model parameters, “including updates on risk weights and correlations”, do not count as changes. A validation once granted then remains in place, and under Article 3(4) no new procedure is triggered. Only an update that touches the model's design, risk capture or assumptions, or that adds or removes asset classes, counts as a change.

That is a pragmatic choice, since a half-yearly revalidation with a six-month decision period would be out of date before it was finished. It does, however, move supervision from a one-off review to ongoing observation. The decision says as much: because the validation is not time-limited, its tenth recital calls for it to be accompanied by “continuous monitoring of the model by the EBA”. Article 7(2) names calibration and performance based on backtesting results as subjects of that monitoring. The reporting framework is the tool for it. It supplies data for exactly the stretch between two structural changes, during which the model runs on without a fresh review.

Anyone who has reconciled margin calls on a trading floor knows that an initial margin model rarely reveals its weaknesses in the methodology document and usually does so in the next morning's dispute. In trading-book terms, disputes and backtesting exceedances are the profit and loss (P&L) explain of a margin model: they show where the calculated number and the actual loss diverge. The draft targets exactly that. Under the proposal, entities with significant OTC activity would report formal margin disputes every quarter, together with the most significant disputes in detail, all add-ons applied to the SIMM result, cases of deliberate non-use, and backtesting results, both aggregated and for individual portfolios, supported by a free-text analysis of root causes.

The British supervisor's finding concerned large banks in the United Kingdom and predates EMIR 3. It nonetheless describes the arrangement at issue in the EU as well: the model's owner calibrates, users rely on that calibration, and supervisors see performance only at one remove. The EU draft addresses the same gap with data at the level of the individual counterparty. Whether that closes the gap or merely documents it depends on the quality of the root-cause analyses that will sit in the free-text fields.

A group's derivatives volume decides the reporting cycle

The tiering is the part of the draft most easily lost in any summary. The consultation paper describes a “two-step filter”. The first step looks at the group's aggregate average notional amount (AANA) of non-centrally cleared derivatives. The line sits at €750bn, the figure EMIR uses for the supervisory procedures under Article 11(15). According to the paper, the filter is explicitly based on the AANA of the group, “but not the type of authorisation or activity”. In the second step, an EU entity must either exceed a monthly volume threshold of “EUR [X] bn” in its own right or be the EU entity with the highest monthly volume in its group. The EBA says it will calibrate X only once it has the 2026 fee data. The qualitative criterion, a footnote adds, “may be dropped”.

Two things follow, one certain and one open. An entity in a group below €750bn AANA already knows that under the draft it reports only once a year, and only two templates: an overview of initial margin as at the December reference date, and the data used to calculate fees. These firms report too, just on a small scale. In large groups, the EU entity with the highest volume reports quarterly for as long as the qualitative criterion stands. What remains open is which other entities in large groups will cross the entity-level threshold into quarterly reporting. This is not confined to banks. The EBA wants the reporting population to cover, ideally, “at least one entity of every group engaged in OTC trading activities in the EU”. Funds, insurers and non-financial companies within large groups may therefore end up reporting quarterly, even though these are not entities that bank supervisors normally oversee. For the threshold test, the paper accepts an estimate, because calculating AANA precisely is “associated with a noteworthy cost”.

Operationally, most of the effort lies in two concepts. The draft introduces the “margined portfolio”, a unit that differs from the netting set. Reports are to be linked to EMIR trade repository data through the collateral portfolio indicator or the unique transaction identifier, a link that question 3 addresses. For backtesting, the proposal envisages data on 30 portfolios per reporting entity, with at least 15 red, 10 yellow and 5 green for groups with an AANA of €750bn or more. Disputes below €500,000 could be left out, as could those resolved within no more than five business days. All of these figures are subject to consultation. The root-cause analysis of exceedances is work for the model risk function and cannot simply be exported from a collateral system.

Washington approves dealer by dealer, London has declined pre-approval

Comparing the other two large derivatives markets shows how far the EU is going. In the United States, supervisors approve initial margin models firm by firm and only for swap dealers. Under section 23.154 of Title 17 of the Code of Federal Regulations (CFR), such a dealer needs written approval from the Commodity Futures Trading Commission (CFTC) or the National Futures Association (NFA) and must give 60 days' notice of material changes. Dealers that are banks follow the same pattern under the rules of the Federal Reserve and the Office of the Comptroller of the Currency. Since October 2021 the NFA has also required quarterly reports, including “a listing of the top five backtesting and benchmarking shortfalls”. Firm by firm, the US regime is in places stricter than the EU draft. Its rules, however, provide for no central validation of the industry model itself. Dealers' clients also remain outside the approval process.

The United Kingdom deliberately chose a different path. In 2023 the Prudential Regulation Authority (PRA) and the Financial Conduct Authority (FCA) proposed “not to implement a supervisory pre-approval requirement at this stage for using initial margin models” and confirmed that decision in December 2023. The PRA reviews SIMM within its existing supervisory framework, as its 2022 letter shows, without a dedicated approval process. Until EMIR 3, the EU had no such process either. In an opinion published on 3 July 2023, the EBA noted that the European mandate, “contrary to the BCBS-IOSCO standards” of the Basel Committee on Banking Supervision and the International Organization of Securities Commissions, did not provide for supervisory approval, and described a “coordination problem”: a model designed in one place but subject to review by many national authorities. In December 2022 the Commission even wanted to delete the existing mandate for supervisory procedures on model validation under Article 11(15)(aa). Central validation entered the text only during the legislative process, and the industry had asked for it. In a letter dated 17 May 2019 to the three European Supervisory Authorities (ESAs), ISDA, SIFMA, SIFMA AMG and ALFI wrote: “It would then seem logical that the ESAs are directly in charge of the validation rather than the NCAs.”

Of the three large markets, only the EU requires approval for every user while also running a central validator for the industry model. The regime is not finished: the regulatory technical standards and guidelines on model approval were out for consultation until June 2026 and have not been adopted. Back in 2023 and 2024, the main points of contention were already the reach of European review beyond the large dealers and the duplication of effort. In a letter dated 3 October 2023 to the Council, Parliament and Commission, ISDA called it “disproportionate and unnecessary to require smaller institutions to obtain initial margin model approval”.

Two monitors count the same disputes differently

The EBA is not the only party watching the model. ISDA has run its own monitoring for years, set out in its governance framework, last revised on 18 September 2026. Users are expected to report quarterly data on shortfalls and on reconciliation issues that led to a dispute “of at least 10 days in duration”. The same framework states: “All data collected from SIMM users for any purpose related to the governance of the SIMM will be anonymized”. The EBA draft visibly draws on this source. Its list of possible causes of margin divergence, the paper says, is “inspired by ISDA's quarterly ISDA SIMM monitoring reports”. Under the February decision, ISDA is itself a user of the EBA system and supplies the documentation for every update there.

The EBA justifies its own collection by saying it found no alternative: “No other sources have been identified that could provide the specific information intended to be collected.” That carries weight, since ISDA's data is anonymised and shows supervisors no individual counterparty. ISDA, however, says it provides model documentation and monitoring results to 22 supervisory authorities worldwide. The paper does not examine whether the EBA is among them, or whether its collection could partly rely on that material. By 18 October no industry association had published a response to the draft. The EBA itself will publish responses only after the deadline. One earlier position is on record: in March 2026 ISDA described reporting initial margin amounts for the fee calculation as “burdensome and unnecessary”, arguing that volumes could be derived from trade repository data.

On the draft as it stands, and without predicting the final text, two data sets would cover the same events if the proposed definitions survive. ISDA counts a dispute once it has lasted ten days, the EBA once it reaches €500,000, with cases lasting no more than five business days allowed to drop out. Timing adds a further layer. ISDA recalibrates every six months, with effect in July and December, while reporting is quarterly, so a version change will regularly fall in the middle of a reporting period. Both are manageable, but they require coordination between collateral operations, model risk and regulatory reporting. Questions 8 and 9 of the consultation ask about exactly these thresholds. Firms that track disputes today only under ISDA's definition would do well to plan for the second count early.

Recommendations for collateral, treasury and model risk

With the deadline close and important points still open in the draft, five steps stand out for firms with SIMM-based models.

Glossary

Christian Schablitzki – the agentic banker

Strategy & Management Consultant · Agentic AI expert for financial institutions

More than 20 years in investment banking and derivatives trading, followed by over 10 years advising financial institutions. Currently a Partner at Infosys Consulting in Germany. Certified as Claude Partner – Claude Code (Anthropic), as well as in Google AI, Generative AI Leader (Google Cloud) and IBM RAG and Agentic AI.

LinkedIn profile → · agenticbanker.ai →