CSSF Circular 24/856, annotated — thresholds, procedures and the FAQ, in one place

People arrive at this circular by googling a paragraph number — "paragraph 121 of circular 24 856", "point 35 tolerance threshold", "24/856 de minimis" — because they are reading a 47-page CSSF text unaided, at the desk, mid-remediation. The law-firm alerts that landed in April and December 2024 are day-one summaries; useful, but they stopped at the circular. The CSSF's own FAQ (Version 1, 24 December 2024) sits alongside the text and answers the operational questions the circular leaves implicit — closed-ended funds, settlement-mismatch breaches, what actually counts as a fee error, whether you need CSSF sign-off before applying de minimis. This page puts the circular and its FAQ in one place, point-numbered, so you can stop scrolling the PDF.

Circular CSSF 24/856 entered into force on 1 January 2025 (pt 162) and repealed Circular 02/77, the 2002 text the industry had memorised, on that date (pt 166). Errors detected between the circular's publication and 1 January 2025 continue to run under 02/77 (pt 163) — the trigger is the detection date, not the error date. Everything below is the new regime — for errors detected on or after 1 January 2025.

The cross-domicile view — how Luxembourg's thresholds compare with Ireland, the UK, Jersey and Guernsey for the same NAV error — lives on the NAV error correction matrix. This page stays inside 24/856.

Scope — which vehicles, and which chapters

The circular applies to UCITS, Part II UCIs, SIFs and SICARs directly, and reaches ELTIFs, MMFs and EuVECAs/EuSEFs through points 2–4 of Section 2.1 (its scope section). The FAQ's first clarification is the one most people need: RAIFs are not in scope — except where a RAIF falls into a listed category for which the CSSF is the competent authority, the worked example being a Luxembourg ELTIF (FAQ Ch. I, Q1). If your RAIF is a plain SIF-regime RAIF, 24/856 does not bind it; if it is set up as an ELTIF, it does.

The reach is not uniform, though, and the difference is one chapter. UCITS, Part II UCIs, SIFs and SICARs get the circular in its entirety (pt 1). The vehicles pulled in through points 2–4 — Luxembourg ELTIFs that are not Part II UCIs/SIFs/SICARs (including RAIF-ELTIFs), MMFs that are not UCITS/Part II/SIFs, and Luxembourg EuVECAs/EuSEFs — get the circular in its entirety except Chapter 8, the réviseur d'entreprises agréé regime. So a RAIF-ELTIF sets thresholds, corrects, compensates and notifies like everyone else, but the separate- and special-report machinery in Chapter 8 does not attach to it.

Two chapters do the heavy lifting, and they switch on separately:

Closed-ended funds are not off the hook, though. The carve-out sits in the last paragraph of pt 27, and it is narrower than "exempt": a closed-ended UCI must still follow pt 26 and the first paragraph of pt 27 — reliable valuation, policies and procedures that limit and detect NAV calculation errors — and must "ensure that corrective measures have been taken to correct, where appropriate, any NAV calculation error". What it does not do is set a mandated threshold or file with the CSSF. And the audit check survives: the same paragraph requires the closed-ended fund's REA to verify, as part of the statutory audit, that the NAV accurately reflects the value of the UCI's assets and liabilities calculated in line with the constitutive documents and/or prospectus — repeated as the footnote to Table 9, where a significant NAV error in a closed-ended UCI is marked "No intervention (*)". The FAQ then goes further and recommends a closed-ended fund define a tolerance threshold internally, purely to identify which errors need correcting — and confirms that doing so does not pull it into the Chapter 9 notification duty (FAQ Ch. II, Q3). So the sensible closed-ended posture is: set a threshold for your own control file, correct above it, notify no one, and expect the auditor to look anyway.

Who owns the error — Chapter 3, and the three routes to the CSSF that bypass you

Chapter 3 is four pages nobody quotes and it settles the two questions every incident opens with: who has to fix this, and who ends up paying. The operative principle is point 15, repeated at point 117 — the party that caused the error, through non-compliance with the obligations applicable to it, must ensure compensation. That is the sentence pt 133's "the fund does not bear the correction costs" rule is built on.

Responsibility for causing an error and responsibility for getting it corrected are deliberately split. Where the UCI has designated an IFM, the IFM — under the supervision of the UCI dirigeants — is responsible for ensuring the correction and remediating the loss to the fund and its investors; where there is no IFM, the UCI dirigeants carry it themselves (pt 16). And appointing an IFM does not hand the problem over: point 17 keeps the UCI dirigeants in charge of the UCI's overall policy "even under a delegation arrangement", requires them to supervise the IFM and permits them to impose rules of conduct on it, including where the IFM in turn sub-delegates. For a fund in contractual form (FCP), the IFM's dirigeants are deemed to be the UCI dirigeants.

Delegation thins nobody's obligations, at any level of the chain:

• IFM — its responsibility "is not affected by the fact that it has delegated any functions to a third party" (pt 19, citing Article 110 of the UCI Law, Article 18 of the AIFM Law and Article 75 of Regulation 231/2013), and the delegations and sub-delegations in place "shall not prevent the IFM from executing its own obligations" (pt 20). Valuation and NAV determination are named as collective-portfolio-management functions the IFM owns.
• UCI administrator — under Circular CSSF 22/811 it must have procedures to detect, assess and notify errors to the UCI or IFM; errors must be recorded, monitored and corrected upon the UCI's or IFM's approval, and the administrator draws up the remediation plan for their approval, implements the corrective measures and follows them up (pt 23). Its own delegation does not exempt it either (pt 24).
• UCI dirigeants — must ensure the relationships with every service provider involved in the fund's functioning are covered by adequate arrangements, "including contractual arrangements" (pt 14). The contract is where "who caused it, who pays for it" gets settled, and the time to settle it is before you need the answer.

Then the part that changes how you handle a borderline call. Three separate parties have a standing route to the CSSF that does not run through you, and all three are triggered by your inaction, not by the error:

• The depositary — where it notices an error it informs the dirigeants of the UCI or IFM; where it notices that those dirigeants do not act on it, "the depositary must notify it to the CSSF" (pt 25).
• The REA (the fund's approved statutory auditor) — same shape: it informs the dirigeants without delay and asks that the error be treated and notified; if they do not follow up it notifies the CSSF at the latest with its management letter and separate report — and immediately where the inaction may adversely affect the rights of the UCI or its investors (pt 140).
• The UCI administrator — under 22/811 (as amended by Circular CSSF 25/900) it must inform the CSSF without undue delay where it becomes aware that the UCI or IFM does not properly fulfil its legal, regulatory or contractual obligations, or does not notify the CSSF as required, of NAV errors, breaches or incidents — material or non-material (22/811, pt 18).

The practical consequence: a notification you decide not to file is not a decision you make alone, and the version of the incident that reaches the CSSF second is worse than the one that reaches it first.

One routine check behind all of this stopped at the end of 2025, and it changes what the CSSF sees between notifications. Until then the UCI administrator's annual return — Annex B of 22/811 — counted NAV errors, material NAV errors and corrected NAVs, each split by whether the administrator was the responsible party, alongside a count of active investment breaches. Circular 25/900 repealed Annex B, and the replacement format the CSSF published on 31 December 2025 (UCI Administrator Annual Reporting — practical and technical guidance, v2.0, with JSON schema v3.0.0) drops those fields: the return now collects the number and net assets of the UCIs administered per function, and what is delegated, with no error field at all. The predecessor schema (v2.0.0, 27 September 2023) carried them explicitly — rgstrNavErrors…, …MaterialNavErrorsForRegLuxUCI…, …CorrectedNav… and navFaActiveInvestBreaches. None survive. Two caveats before you read too much into it: this is verified for the UCIAR channel, and for the entity types whose UCIA reporting was folded into another circular's long-form report or self-assessment questionnaire — 22/821, 23/839, 24/850 and 25/870 for credit institutions, IFMs, support PFS and investment firms; the Specialised PFS SAQ for specialised PFS — the question sets are not published, so we cannot say the same (CSSF communiqué, 16 December 2025; entities outside those two channels still file through UCIAR, at the latest five months after their financial year-end). The practical read: no routine count now tells the CSSF that your NAV errors are accumulating. What does is your own 24/856 notifications, the three escalation routes above, and the auditor's sampled procedures — which is worth knowing before treating a borderline error as a filing you can quietly absorb.

What counts as a NAV error at all (pts 28–30)

Before the threshold question there is a prior one the circular answers in a single sentence, and it is the most useful sentence in Chapter 4 for anyone defending a valuation after the fact. Point 29: the NAV is presumed to be correctly calculated where the rules in the law, the constitutive documents and the prospectus were "applied consistently and in good faith, on the basis of up-to-date and reliable information available at the time of the calculation." The test is the information you held then, not the information you hold now. A price that later proves to have been wrong is not, by itself, a NAV error — which is why the FAQ's audit-fee example (an accrual booked in good faith on reliable information that later proves too low) is not treated as an error at all.

Point 30 comes at it from the other side and lists what does cause one: human error, inadequate control procedures, shortcomings in the administrative processing of operations, imperfections or deficiencies in the IT, accounting or communication systems, and non-compliance with the valuation rules in the law or the fund's own documents. It also places them — these factors "may arise at the level of the entity which calculates the NAV (in principle, the UCI administrator) but also at other levels of the UCI organisation (e.g., dirigeants of UCIs, IFMs, external valuer, etc.)". The administrator is where errors are usually found; it is not where they are all made, and pt 15 allocates the compensation by cause, not by who ran the calculation.

The tolerance thresholds (pt 35)

The threshold is the error size — as a % of NAV — above which the error is significant and the full correction-and-notification machinery switches on. Below it, you fix the cause and move on. The grid is prescribed for retail vehicles and self-set (within a hard cap) for professional ones.

The self-set threshold in pt 35(d) is where an inspection starts, because the circular says what the analysis must weigh and then reserves the right to read it. Five factors are mandatory — the text says "at least" — under sub-point (i): the UCI's characteristics including whether it is open- or closed-ended and how often it opens for subscriptions and redemptions; the investment policy set out in the constitutive documents and/or prospectus; the nature of the investments (listed versus unlisted is the circular's own example); the risk profile, naming liquidity, market and credit risk and the fund's volatility level; and the valuation policy in place. Table 3 is a reference point, not a floor to walk away from — sub-point (ii) says the CSSF expects a professional-investor Part II UCI, ELTIF or SIF running a policy similar to a Table 3 line to land on a similar number. Sub-point (iii) carries the 5% cap, the ban on using 5% by default, and the part people miss: the UCI "must make available to the CSSF" the specific analysis, and the CSSF reserves the right to request it and ask for more. A threshold set above Table 3 then triggers disclosure — investors must be informed through the official and usual communication channels named in the constitutive documents or prospectus; the same condition attaches to a retail Part II UCI or ELTIF using the pt 35(c) derogation to go past 1%, which must also meet sub-points (i) and (iii) of (d). A self-set threshold with no written analysis behind it is not a threshold, it is a finding waiting for an on-site.

The mechanics that trip people up, all in pt 37 and pt 38:

• One threshold per sub-fund, across all share classes (pt 37(i)) — you do not set a looser threshold for an institutional class inside a retail sub-fund.
• Look-through for funds-of-funds, index trackers and feeders — the threshold keys off the target exposure, not the wrapper (pt 37(iv)).
• The cross-border ratchet — where a country in which the units are admitted to marketing requires a stricter threshold, the UCI "must retain this more restrictive tolerance threshold", and it then applies to the whole sub-fund and every investor in it, "regardless of their country of origin" (pt 37(v)). It is not an election: a single foreign registration can tighten your Luxembourg tolerance for everyone.
• Aggregation — sub-threshold errors that run simultaneously or in succession are added together; if they stack past the threshold, they count as one significant error (pt 38). An error log that looks clean line-by-line can still be reportable.
• A lower internal threshold binds you — pt 36 expressly permits a UCI to apply thresholds lower than Tables 1–3, and the FAQ closes the obvious temptation: once your own approved policy sets a threshold below the circular's, all of 24/856 (correction and notification) keys off your number. You cannot set 0.30% internally and then notify only at 0.50% (pt 36; FAQ Ch. II, Q2).

Active vs passive non-compliance — the gate before Chapter 5

Chapter 5 governs breaches of the investment rules, and almost everything in it hangs on one classification made at the moment of discovery. Point 54 requires the UCI or IFM to determine whether a breach is active or passive. Point 55 defines passive narrowly: breaches that "occur for reasons beyond the control of the UCI or as a result of the exercise of subscription rights" — the worked example being market-price movements that shrink the NAV a restriction is measured against. Active non-compliance is the residual: everything that "cannot be qualified as being beyond the control" of the UCI. Point 56 spells out what that catches — voluntary investment or divestment decisions, and the absence of an action or decision "where non-compliance with an investment rule was foreseeable or avoidable" — arising from "an intentional decision, inadvertence, human error or from operational or technical/IT failures", whether at the dirigeants, the IFM or a delegate.

The classification is the whole game, because point 55 puts passive non-compliance outside Sections 5.4, 5.5 and 5.6 entirely and states that it "must not be notified to the CSSF in accordance with Chapter 9". No correction machinery, no filing. What you owe instead is corrective measures "within a reasonable period of time", with the position monitored by the dirigeants until it is cleared — and for non-UCITS holding less liquid assets (SIFs, SICARs, Part II UCIs, open- or closed-ended), point 55 expressly allows the dirigeants, "in specific and duly justified circumstances", to keep the offending position where investors' interests justify it.

Which is why the FAQ's third chapter matters more than its length suggests: it spends four of its twelve questions pushing back on optimistic "passive" calls. A settlement mismatch that pushes deposits over the 20% limit is active, because it was predictable (FAQ Ch. III, Q1); a shortened US settlement cycle is something your pre-trade controls were supposed to model, not an act of God (Q2). Treat "passive" as a conclusion you must be able to defend in writing, not the default for anything you did not intend.

Which rulebook the breach is tested against (pt 49) — and the reference that has since moved

Point 49 defines the perimeter: all asset-eligibility rules, portfolio-management techniques and investment restrictions in the regulatory texts — European regulations, sector laws, Grand-ducal and CSSF regulations, and CSSF circulars — plus, under point 50, the contractual rules in the fund's own constitutive documents and prospectus. Point 51 is blunt about the consequence: "all the investment rules that apply to a UCI are to be considered". A prospectus limit you wrote yourself is an investment rule for Chapter 5 purposes, and breaching it is the same event as breaching a statutory one.

One of point 49's named examples has since been repealed. It cites "the investment restrictions laid down in Circular CSSF 07/309 on risk-spreading in the context of specialised investment funds" — and 07/309 was repealed on 19 December 2025 by Circular CSSF 25/901 (pt 51), together with Circulars 02/80 and 06/241 and Chapters G and I of IML 91/75. 24/856 has not been amended to follow, so its cross-reference now points at an archived text. In practice the perimeter is unchanged in kind but changed in content: 25/901 sets risk-spreading by investor type instead — 25% per position, raised to 50% for funds reserved to well-informed or professional investors (25/901 pts 8 and 11) — while grandfathering the rules of funds and compartments authorised before 19 December 2025, which may continue to apply them (pt 50). So for a legacy SIF the old 07/309 spread may still be the rule your breach is measured against; for anything authorised since, it is 25/901's. Get that wrong and the breach test in pt 91 runs against the wrong number. The full picture is on the Circular 25/901 page.

That clearing-out has since finished, and it took a text the perimeter had leaned on for thirty-five years. The chapters of Circular IML 91/75 still standing after 25/901 took Chapters G and I were repealed in full on 22 May 2026 by Circular CSSF 26/912 — a two-page circular that repeals "with immediate effect", carries no transitional provision and puts no replacement text in their place. What went was anchored to the superseded Law of 30 March 1988 and had long been read through its successors, but until that date it sat inside pt 49's "CSSF circulars" perimeter: Chapter F, the UCITS rules on asset composition, ancillary liquid assets and borrowings — and, next to pt 52's calculation basis, the sentence that "the investment limit percentages to be complied with by UCITS must be applied to the net assets of UCITS"; Chapter H, the techniques and instruments for efficient portfolio management (options, financial futures, securities lending, repurchase agreements, and currency-risk hedging), which 25/901 had already stopped applying to Part II UCIs; and Chapter M, whose monthly reporting required an explanation wherever the NAV per unit moved more than 10% against the previous month-end — the last NAV-anomaly trigger sitting at circular level outside 24/856 itself.

The CSSF's stated reason is that the content is either overtaken by national and EU rules or "incorporated into the CSSF's administrative practice". That second half is the part to carry into a control file: a rule surviving only as administrative practice is not a published text you can cite when you are evidencing which limit a breach was measured against. Where an internal procedure or a prospectus still points at IML 91/75 by name, the citation is now dead — and under pt 50 a rule you wrote into your own constitutive documents remains an investment rule for Chapter 5 whether or not the circular behind it still exists.

Two further sentences in the same section decide what the limit is measured against, and for how long. Point 52 — the calculation basis. For UCITS, the quantitative investment rules in the regulatory framework apply to the NAV; where the limit is contractual, the constitutive documents and/or prospectus "must specify the calculation basis", general practice being the NAV again. For non-UCITS the regulatory texts themselves use different bases — the circular names the NAV and the capital of the UCI — and contractual limits are set in practice against the NAV, capital or gross assets. A prospectus limit drafted without a stated basis is a breach test you cannot run, and pt 52 puts fixing that in the fund documents, not in the control file. Point 53 — the time perimeter. Compliance is owed on an ongoing basis through the life of the UCI until dissolution or liquidation, and the only exceptions are those a regulatory text or the fund's own documents provide. The circular names two: the six-month period in Article 49(1) of the UCI Law, which lets a newly authorised UCITS derogate from the investment limits of Articles 43 to 46 for six months from authorisation; and the longer ramp-up and divestment periods AIFs write into their own documents to build a portfolio or sell it down ahead of liquidation. Inside a documented window there is no non-compliance to classify at all — which is why the window has to be in the prospectus before you need it. Outside one, "we were still building the portfolio" is not an argument, it is a missing clause.

The controls that decide whether you ever see the breach (pts 57–64)

Section 5.3 is the part of Chapter 5 nobody quotes, and it is the one your control file gets judged against after the fact. Point 58 splits the arrangements in two. Pre-trade controls test the intended transaction against the applicable rules at the moment of the portfolio-management decision — at the IFM or its delegate — and are "all the more important" for UCIs that may hold less liquid or illiquid assets. Post-trade controls run after the trade, and the circular gives them a specific job: determine whether the non-compliance is active or passive, so as to trigger the corrective action and the impact calculation. Post-trade also carries the only hard timing in the section — it "must occur at the latest when calculating the next NAV". Frequency and design go in the written procedures (pt 59).

Then the principle that catches funds with infrequent valuation points. Point 60 states it flatly: the UCI must comply with the applicable investment rules at all times — at every NAV and between two NAVs, including on an intra-day basis. Point 61 pushes that into the controls: they must detect, correct and notify transactions carried out between two NAVs that breach the eligibility rules or the same-issuer holding limits. Points 62–64 then set the proportionality most funds actually live under:

• The general concession (pt 62) — for diversification and other restrictions measured against net assets (or another reference the regulation or prospectus uses, gross assets being the circular's example), the CSSF considers it generally sufficient to verify at each NAV calculation. That is the default an administrator will be running.
• The exception that removes it (pt 63) — where a UCI is materially exposed to between-NAV breach risk because it engages significantly in transactions between two NAVs, including intra-day, it must have controls proportionate to the nature, frequency and complexity of those transactions to ensure compliance between the NAVs. An actively trading fund cannot lean on pt 62.
• Infrequent NAVs (pt 64) — a monthly or quarterly striker that transacts between NAVs must verify by another route, the circular's own example being a control at the time of the transaction. A long gap between valuation points is not a defence.

The link back to the active/passive gate is direct, and it is why this section is worth reading before the exciting chapters: if pt 63 described your fund and you were only checking at each NAV, the breach was foreseeable and avoidable — which is pt 56's definition of active.

Correcting an active breach — the accounting method by default, and the price of the economic one

Once a breach is classified active, Chapter 5 splits into two questions that get conflated at the desk: what you do to the portfolio, and what you owe the fund. Point 67 lists the corrective actions by example — sell the ineligible investments, sell the excess positions, adjust a portfolio that sits under a prospectus minimum, cut borrowings back below the cap, or have the counterparty post eligible collateral on a securities-financing transaction — and 67(vi) concedes that sometimes no corrective action is needed, because market movement or capital activity has already cured the position. Point 69 attaches a different impact test to each: sales and portfolio adjustments require a full impact calculation; a collateralised securities-financing breach needs none unless the counterparty defaulted during the breach period; and a self-curing breach still needs the unrealised result computed from the irregular investment's prices across the period. "It fixed itself" is not "nothing to calculate".

The clock on that correction runs at two speeds, and only one of them stretches. Point 66 requires the necessary measures to be decided without delay in every case; what varies is execution. For UCITS and UCIs invested in liquid assets, the corrective action "must be implemented directly after the detection" of the active non-compliance. For non-UCITS holding less liquid or illiquid assets — the circular names SIFs, SICARs and Part II UCIs — the decision is still owed without delay, but the circular accepts that implementing it "could possibly take longer", scaled to the fund type, its investment policy and the assets held. Point 65 sits in front of both: on discovery, the procedures must inform the IFM, the UCI administrator and the depositary without delay, and the UCI and IFM dirigeants must be informed well enough to perform their Chapter 3 role. Illiquidity buys you execution time, not decision time — and a dated decision record is the only thing that evidences which of the two you actually used.

Then the line that inverts most people's instinct: the NAV tolerance thresholds cannot be applied here (pt 71). Money owed to the fund for a breach has no materiality filter — the threshold only ever governs the separate NAV-error leg under pt 91. Timing is prescribed too (pt 76): for portfolio transactions the impact period runs trade date to trade date — the trade that caused the breach to the regularisation trade (or the date compliance returned unaided); for cash and deposit breaches it runs on value dates, effective inflow to effective outflow. And if the calculation nets out to a gain, the gain stays with the fund (pt 75) — you do not net it against the next loss.

The impact itself runs on one of two methods, and they are not a free choice:

Three constraints sit on top of both. The method must be defined in an internal policy from launch, precise enough that instances are treated consistently and that nobody can arbitrage the method against the compensation number it produces (pt 72) — an umbrella may set it fund-wide or per sub-fund, and may vary it by breach type (pt 73). Switching method after the fact is barred: a change from accounting to economic must be justified in investors' interests and approved by the dirigeants, and "in principle" may not be applied to a breach already identified (pt 74). You cannot run both calculations and file the cheaper one. And where several breaches are simultaneous — closely linked, same origin or nature, overlapping periods — you may net the regularisation results across them (pt 87), but you must be able to show the netting cost neither the fund nor its investors anything (pt 88). Breaches that fail that closeness test are calculated one by one (pt 86).

The FAQ then tests exactly this machinery twice, and both answers point the same way: you may not net negative deposit interest between banks — the fund is indemnified for the interest and charges it actually bore, on the accounting method, unless the §5.5.3.2 conditions above are met (FAQ Ch. III, Q4); and you may mix methods within one UCI only where the policy says in advance which method attaches to which breach type (Q5). Both are policy questions answered before the breach, not calculation questions answered after it.

When a breach becomes a NAV error too

One route into Chapter 4 gets missed because it starts somewhere else. An investment-rule breach becomes a significant NAV error in its own right where capital activity — subscriptions or redemptions — was executed at an erroneous NAV during the breach period (pt 90). The circular puts the test on you: for every instance of active non-compliance you must check the impact on all NAVs struck during the breach period against that sub-fund's tolerance threshold (pt 91), and where it is exceeded, recalculate and correct the NAV for each day it was exceeded under Chapter 4 (pt 92). A breach remediated in the portfolio is not finished until that test is run. The same doubling-up runs through the "other errors" in Chapter 6 — swing-pricing errors (Tables 6–8) and cut-off errors, where pt 110 makes you verify that your correction did not itself produce a significant NAV error.

Fee errors work the other way round, and this is the trap. Where a UCI paid too much in costs or fees, pt 104 says the charging "is an error which must be corrected and for which the UCI must be compensated" — and then, flatly, that "the tolerance thresholds provided for in Chapter 4 do not apply in this case". There is no materiality filter on an overcharge: the full amount goes back, at any size, and only if that compensation itself causes a significant NAV error does Chapter 4 switch on. Underpaid fees (pt 105) fork two ways — either the party that caused the error pays the shortfall from its own assets, or the UCI recovers it retroactively from fund assets, in which case you must correct the erroneous NAVs across the whole error period "without applying the tolerance threshold", and ensure the cost lands only on the investors who actually got the service. Either way, "it's below our threshold" is not an answer to a fee error.

Chapter 6 — the other four error types, and the catch-all

Chapter 6 covers what is neither a NAV calculation error nor an investment-rule breach, and it is where four of the six notification form types come from: swing pricing, non-compliant payment of costs/fees, cut-off, and investment allocation (pt 94). Two framing points get skipped. Point 95 fixes the escalation — on discovery, the procedures must inform the IFM, the UCI administrator and the depositary without delay, and the UCI and IFM dirigeants must be informed well enough to perform their Chapter 3 role. Point 96 is the catch-all: for other errors in the UCI's activities and operations, the absence of specific guidelines in the circular removes nothing — "at every occurrence" the UCI must assess whether corrective actions and compensation for the UCI and/or investors are required. There is no fifth named category to hide in.

The four named types diverge on the question that decides everything at the desk — does the tolerance threshold filter this error at all?

Swing pricing — and every other dilution tool (pts 97–100)

Three tables, three failure modes, and they do not all cost the fund money. Table 6 (the swing factor was not applied at all): the fund was unprotected from dilution and its loss is the net variation in capital multiplied by the difference between the NAV applied and the NAV that should have been applied — the fund is compensated. Table 8 (the factor applied was insufficient) works the same way on the shortfall. Table 7 is the one people get wrong: where the factor applied exceeded the one set for the fund, the fund was over-protected, suffered no loss, and no compensation to the fund is required. In all three, some investors gained and others lost depending on the direction of the swing, and the investor leg runs the same test: below the tolerance threshold, no compensation for capital activity; above it, the full Chapter 4 correction.

Point 100 is the sentence that widens the section past its title. The same guidelines apply "in a similar manner" to errors in other dilution-management tools, taking each tool's specifics into account — the circular names the anti-dilution levy, liquidity fees reflecting the cost of the fund providing liquidity, and commissions charging investing or divesting investors the transaction costs, taxes or other elements, whether applied as a lump sum or calculated individually. A fund that runs an ADL rather than swing pricing is inside §6.1 regardless of what its prospectus calls the mechanism.

Costs and fees — the line pt 102 draws (pts 101–105)

Section 6.2 has a boundary that decides which regime you are even in, and it is not about size. Point 102 carves out costs/fees that were merely provisioned and never effectively paid by the UCI: those are outside Section 6.2 entirely, and a significant provisioning error is treated as a plain significant NAV calculation error under Chapter 4. Section 6.2 bites once money has actually left the fund. That single distinction is what drives the FAQ's three worked examples (Ch. IV, Q1) — the accrual booked in good faith on reliable information is not a fee error; the wrong accrual caught before payment engages neither Section 6.2 nor Chapter 4 if no significant NAV error resulted; the wrong accrual that was paid out does engage Section 6.2. Check "was it paid?" before you check anything else.

Cut-off — and the second regime sitting behind it (pts 106–111)

A cut-off failure means orders submitted in due form were executed on a NAV earlier or later than the one the constitutive documents and prospectus entitled the investor to (pt 107). Correction under pt 108 covers all the orders executed on the erroneous NAV, and splits the same way as a NAV error: investors who realised a gain keep it, with the fund compensated for the difference between the erroneous and corrected NAV; investors who lost are compensated by the fund, either through the allocation of additional units or by paying the difference. Clawback follows the familiar line — a well-informed or professional investor may be asked to repay the amounts unduly received; for anyone else it is "in principle not appropriate".

Two checks then run on the correction itself. Point 110 makes you verify that the corrective actions did not themselves produce a significant NAV calculation error requiring Chapter 4 treatment. And point 111 is the one that turns a cut-off incident into a second file: the UCI must also verify whether the incorrect application of the cut-off entailed non-compliance with Circular CSSF 04/146 on the protection of UCIs and their investors against late trading and market timing — and if so, that circular's provisions apply as well. A late order accepted at a stale NAV is a cut-off error under 24/856 and potentially a late-trading matter under 04/146; the two are not alternatives.

Investment allocation errors (pts 112–115)

The last named type is the one with no threshold and the least written about it. Point 113 gives three worked shapes: a security that should have gone to sub-fund A of a UCI booked to sub-fund B; a security that should have gone to UCI X booked to UCI Y; and transactions — the circular's example is currency hedging — allocated to the wrong share class.

Point 114 explains why that is a loss rather than a bookkeeping tidy-up, and it names two separate harms. The fund is exposed to the risks of an investment it should never have held from the moment the position is recorded for its benefit in the administrator's books or in its accounts at the depositary. And, on the other side, the fund does not get the performance of the investment the manager intended for it, because that investment went to a third party. Point 115 then sets the remedy: the UCI must be compensated for the loss suffered on the erroneously allocated investments; where those investments generated a profit, the fund keeps it; and where the allocation error caused a significant NAV calculation error, Chapter 4 runs on top. A share-class hedging misallocation corrected quietly in the books, with no impact calculation and no threshold test on the NAVs struck in between, is an incomplete correction.

The notification procedure — form, and the 4–8 week window

Every significant error goes to the CSSF on its prescribed notification form. The form is filed in full — all fields, the quantitative impact calculation attached — and an incomplete filing is refused, so a half-finished form does not stop your clock.

There are exactly two permitted transmission channels, and email is not one of them: initial submission and follow-up through the dedicated eDesk procedure, or automated submission via API on the S3 protocol with the follow-up still done in eDesk. Both went live in production on 1 January 2025, after a familiarisation period on eDesk PREPROD (CSSF communiqué, 17 December 2024). The exclusivity is the CSSF's own word: its procedure communiqué of 31 December 2024 states that any detected error or instance of non-compliance "need to be notified exclusively through the two transmission methods" above. A PDF emailed to your usual CSSF contact is not a notification. Where the form cannot carry everything needed to understand the error, the additional explanations go in a separate document sent with the form (pt 156) — not as a later supplement.

"The form" is really six typed forms, and you pick one per incident: NAV calculation errors · non-compliance with investment rules · and four flavours of "other error" — incorrect swing pricing / anti-dilution tool, non-compliant payment of costs/fees, incorrect cut-off application, and investment allocation errors (practical guide §2.1; each carries its own field set in Annexes I–VI, and its own JSON section for API filers). The unit of filing is narrower than most people assume: one notification covers one sub-fund of one UCI — the JSON report is defined that way, and the eDesk form makes you select fund and sub-fund before it opens (standalone funds file under "sub-fund 0"). An umbrella that struck the same wrong FX rate across five sub-funds files five notifications, not one, and each needs its own impact calculation. Budget the time accordingly.

The deadline is 4 to 8 weeks from the date of detection (not from the error date, and not from when compensation finishes). Read pt 158 closely, because it is drafted oddly: the 4–8 week window appears in the sentence about an error "which does not entail the payment of compensation to investors", and complex cross-border compensation cases are then told to report "within the above deadlines" without necessarily naming the investor-payment date. So the window is universal; only the payment date is allowed to arrive late, through the completed form, with monthly progress updates to the CSSF while payouts run. The CSSF's own filing guide removes any doubt and states the expectation flatly: a complete notification "in principle, within 4 to 8 weeks of the detection of the incident".

Pre-notification vs complete notification — the escape hatch, and its price

The form has two modes, and this is the part the circular does not spell out. The CSSF's practical guide — UCI notification in accordance with Circular CSSF 24/856, v1.2, 16 July 2025 — sets out a pre-notification for cases where the recalculation and compensation work genuinely cannot be finished inside the window. It is not a soft option:

• Pre-notification "can only be used on an exceptional basis in duly justified cases" where the correction is "particularly complex and time-consuming". You must describe the error's typology, duration and estimated impact and justify why it cannot be finalised in 4–8 weeks.
• The CSSF can refuse it. A filing that lands in the "Pre-notification refused" status means the justification did not qualify — and you must then create a new complete notification from scratch, with the clock having run. A pre-notification accepted moves to "Pre-notification to be finalised", and you complete the same notification rather than starting again.
• The impact calculation sheet, in Excel, is mandatory for a complete notification (other supporting documents go in as PDF). This is the pt 154 quantitative impact calculation in practice.

What the NAV-error form actually asks for is worth knowing before you open it: the tolerance threshold applied (validated as a figure between 0% and 5% — the pt 35(d) cap is enforced in the form itself); the detection date and the error period; a separate "period of significant impact" start and end, where alternating significant and non-significant days are reported as one outer span rather than day by day; the nature of the error from a fixed list (securities valuation · accounting · fees and accruals · corporate action · derivatives valuation · other); the maximum impact as a % of NAV; the amounts owed to the sub-fund, to investors in cash, and to investors in units, separately; whether a de minimis was applied, and at what level and in which currency; and the payment dates. Whether you used the compound or non-compound method (pt 47) has no dedicated field — the guide asks you to state it in the free-text "other information" box.

Access is the quiet blocker: the filer needs an eDesk account with LuxTrust authentication, linked to the notifying entity. The fund, the IFM and the UCI administrator can each create notifications, and none can edit a notification submitted by another entity. A fund or IFM may delegate creation and submission to a named person at another entity — the invitation expires in 72 hours if not accepted — but the UCI administrator cannot delegate. On a change of IFM or administrator, neither the outgoing nor the incoming party can see the other's filings for that fund, so the notification history does not travel with the mandate. Keep your own copies (the form exports to PDF at any status).

Three more procedural points worth an ops-checklist line:

• Notify even when nobody was paid. A significant NAV error with no subscriptions or redemptions in the error period — so no compensation due to investors or the fund — still requires a notification. The form must carry the source of the error and the corrective measures taken to stop it recurring (FAQ Ch. II, Q1). "No money moved" is not a reason to skip the filing.
• The auditor's special report, where one is triggered, is due to the CSSF within 3 months of the complete notification (pt 157); if it slips, the CSSF must be updated on the report's progress monthly, the same cadence as compensation payments (pt 158). It is filed back into the same eDesk notification as a new version, not as a fresh submission. What triggers it, and what it checks, is the auditor's leg below.
• The file stays open after you submit. A complete notification can come back with CSSF comments ("Actions required"); each comment topic must be answered before any new version can be sent, and the named contact person — by default whoever created the notification — receives the emails. Change of staff mid-remediation means changing the contact on the notification, or the CSSF's questions land in an unwatched inbox.
• The CSSF does not pre-approve your remediation — and silence is not clearance. Point 159 says the corrective actions set out in the form "are not subject to the CSSF's approval" and that the CSSF "will not provide specific feedback" on a notification; filings feed its risk-based supervisory work instead. It then reserves the right to act ex post (pt 160): request more information, require different remedial actions where yours are off-guideline, or demand organisational fixes where the notifications point to deficiencies. Host-country regulators of any market where the fund is distributed must also be informed under their local rules (pt 161).

Investor compensation, de minimis, and who bears the cost

Recalculate the NAV for every date in the error period; compensation is compulsory only for the dates on which the error was actually significant (pt 39). Direction matters: an undervalued NAV means you compensate redeeming investors and the fund; an overvalued NAV means subscribing investors and the fund (Tables 4–5). The financial impact runs on a compound or non-compound method — pick one in your procedures and apply it consistently (pt 47).

Point 40 turns that into the remedial action plan the CSSF expects to see described in the notification form, and it is usable as a checklist in the order the circular sets it out: identify the origin precisely and correct the source immediately, so the next NAV is right; determine the corrected NAVs across the error period; apply them to the subscriptions and redemptions executed on erroneous NAVs to size what is owed to the fund and to investors; book the resulting payments to be received and made in the fund's accounting, without delay, as soon as those numbers are final; inform the investors being compensated, including the modalities; pay; and decide and implement the control fix — the circular's words are "adjustment or strengthening of internal controls" — that stops it recurring. Stopping the bleed comes first; sizing the loss comes second.

Who funds the payout is a separate question from who owes it, and pt 43 draws the line tightly. Investors may be compensated out of the fund's own assets only where the amounts due correspond to excess amounts already sitting in those assets, so that paying them cannot affect the other investors' interests; otherwise the IFM, the UCI administrator or another party involved in the fund's functioning takes over the payment. Two mechanics sit alongside it. Where one investor both gained and lost across several transactions during the error period, the final compensation for that investor may be determined on a net basis — but only where you can establish it is the same final beneficiary through the intermediation chain (pt 45). And whatever investors kept and was not recovered, the fund must still be made whole for (pt 46): the amounts unduly received and not clawed back are compensated to the UCI.

The compound/non-compound choice is worth defining, because the notification form has no field for it and the CSSF's guide asks you to state it in free text. Under the compound method, each NAV calculated during the error period is corrected for both the direct effect of the error and the indirect effect of the cumulative subscriptions and redemptions that were themselves executed at erroneous NAVs. The non-compound method ignores those cumulative effects. Both are permitted; the method must be set out in the fund's internal policies and procedures, and where both may be used, the policy must specify the circumstances for each so cases are treated consistently (pt 47).

De minimis — a cap on the payout, not on the error

A de minimis amount is a per-investor floor below which a compensation payment is not made — it caps the payout, it does not change whether the error was significant. Under 24/856 it is a lump sum reflecting the actual bank charges of transferring the compensation, documented per fund or sub-fund (pts 124–127). Two hard limits on it:

• It applies to investor payouts only — never to amounts owed to the fund. Money owed back to the fund is paid down to the last unit of currency (pts 124–127).
• Compensation delivered by allocating extra units rather than cash skips de minimis entirely (pts 128–129).

The FAQ adds the operational detail the circular leaves out: you do not need CSSF approval before applying de minimis (FAQ Ch. V, Q1). But the CSSF can ask you to justify the level ex post and to evidence that it genuinely represents the bank charges — and it flags EUR 25 as the level above which it will expect that documentary evidence. Your IFM's internal policy must set out the use and the level of de minimis for NAV errors; an undocumented de minimis is a finding waiting to happen.

Paying through the distribution chain (pts 120–122)

This is the passage people arrive here searching for by number, and it is the one that turns a tidy compensation calculation into a project. Most end-investors are not in the register — the financial intermediary is, on their behalf, often with capital activity aggregated across several investors up the chain (pt 120). Point 121 says the UCI must nonetheless ensure the final beneficiaries receive what they are owed, and must have arrangements in place to trace back the intermediation chain so that compensation reflects each beneficiary's individual position and actual loss.

Point 122 is the fallback for when that is impossible: hand the intermediaries the full data set — error period start and end, erroneous and corrected NAV per day, subscriptions and redemptions per day, impact per day — so they can compensate their own clients, and tell investors clearly that final beneficiaries' rights may be affected when compensation is paid where they subscribed through an intermediary. For any UCI that must issue a prospectus, that warning belongs in the prospectus, at the next update (pt 164) — UCITS, Part II UCIs, MMFs and ELTIFs, and SIFs and SICARs too where they still issue units to new investors. A fund with no prospectus obligation, or one that has not yet updated, does not escape the duty: pt 164 requires the same information to go to investors "via the official and usual communication channels" named in the constitutive documents or prospectus. Two related constraints sit alongside: compensation may not be paid in instalments, nor deducted from future remuneration owed to the party responsible for the error (pt 118); and where compensation cannot be paid without delay, the dirigeants must tell the CSSF why, through the annexes to the notification form (pt 119).

Clawback, orphaned amounts, and the sting

• Clawback from investors who accidentally gained is acceptable for well-informed/professional investors, but "in principle not appropriate" for retail (pt 44).
• Unclaimable amounts owed to investors who have since left go to the Caisse de Consignation — formally the State Treasury's Consignment Office (Trésorerie de l'État, Caisse de consignation), Luxembourg's deposit office for unclaimed funds — by a consignment request under Article 1(2) of the Law of 29 April 1999 on consignments to the State, and only after "all necessary efforts", additional research included, have failed (pt 132).
• The sting, and the line people miss: the costs of running the correction cannot be charged to the fund (pt 133). The party that caused the error carries the remediation cost — not the NAV. And the circular closes the two obvious workarounds: "direct and/or indirect" costs are caught, with staff time at the IFM/UCI, legal fees and the auditor's special report named explicitly (pt 134), and the costs may not be recharged indirectly through an increase in the fees the fund bears (pt 135). This is the single most expensive sentence in the circular.

The auditor's leg — what the REA checks, and when (Chapter 8)

Every error in scope also lands in front of the fund's réviseur d'entreprises agréé (REA) — its approved statutory auditor — by one of two routes, and which route you are on decides how fast you have to pick up the phone.

The power behind Chapter 8 is statutory, not circular-made, which is why it is not negotiable in an engagement letter: Article 154(3) of the UCI Law, Article 55(3) of the SIF Law and Article 27(3) of the SICAR Law each let the CSSF require a UCI's REA to carry out a control on one or several particular aspects of the fund's activities and operations. Points 136–137 are the CSSF exercising that power standing, for every fund facing an error or instance of non-compliance in scope of the circular, with the controls performed when the REA draws up its 21/790 separate report.

The default route is the separate report the REA already produces annually under Circular CSSF 21/790: the errors reported to it during the financial year are reviewed on a sample basis, risk-based, against procedures the CSSF sets directly in that report (pt 141). For those cases the information must reach the REA "at the latest" when it starts its year-end work (pt 139) — end of the audit cycle is soon enough.

Point 165 promised those 21/790 procedures would be rewritten to take 24/856 into account, and they were — not by a new circular, but inside the eDesk report templates, with the CSSF setting out the changes in a communiqué of 3 February 2025, for year-ends from 31 January 2025. The direction of travel is a loosening, and it is quantified: the separate report's review of all errors and instances of non-compliance was replaced by a sample-based review, with the sample fixed by vehicle.

The self-assessment questionnaire moved the same way. Its "Investment compliance" section lost the question asking for the number of active instances of non-compliance and gained one on the appropriate classification of the passive ones; its "NAV determination" section was widened to take in the extension of 24/856's scope to Chapter 6's other error types. Read the two together and the supervisory attention has shifted off the count of errors and onto the active/passive call — the judgement the fund makes to keep a breach away from the CSSF is now the thing both the questionnaire and the auditor's sample are pointed at. A single sampled item for a SIF is not a light touch if it is the one you classified optimistically.

The exception is the special report, and it runs on a different clock. It is triggered where both conditions in pt 138 are met: the error or breach relates to a UCITS or a Part II UCI, and the total compensation exceeds EUR 50,000 or the payout to a single investor exceeds EUR 5,000 — with compensation paid to the fund and to all investors counted together for the EUR 50,000 test. The control "must be performed immediately after the detection of the error", the report goes to the UCI dirigeants and the CSSF, and the fund or IFM must ensure the REA is informed "without delay" and given everything it needs. The trigger is detection, not filing — so on a large error the auditor call belongs in week one next to the notification work, not after the form goes in.

Two asymmetries in those tables are worth pinning. For investment-rule breaches, the EUR 5,000 single-investor trigger only bites where the breach exceeded the NAV tolerance threshold — the footnote to Table 10, and the only place the threshold re-enters the auditor question. And passive non-compliance, which never reaches the CSSF at all (pt 55), still reaches the auditor: Table 10 puts it in the separate report on a sample basis for UCITS, Part II, SIFs and SICARs alike, and pt 143 spells out that the REA samples "the adequate classification of instances of 'passive' non-compliance". The call you made to avoid a filing is the call that gets tested.

What the separate-report checks cover is unsurprising — investors identified from the register, NAVs corrected across the error period, amounts owed derived from the corrected NAVs, payment actually made (pt 142); for breaches, the gain/loss calculation, the consistency of your impact method with both the circular and your own internal policy, payment, and whether the breach also produced a significant NAV error (pt 143). It is the special report's five additional checks (pt 147) that read like a list of this page's expensive paragraphs:

• compliance with the Chapter 7 financial-intermediary guidelines when compensating investors (the pts 120–122 trace-back);
• verification that the fees and costs of the correction were not charged or recharged to the fund — pt 133 and pt 135, audited rather than self-certified;
• assessment of whether the de minimis rule was adequately applied (its level, and the bank charges behind it);
• verification that the notification to the CSSF was adequate, "including the exhaustiveness of the content";
• verification that a root-cause analysis was performed and corrective measures taken.

Note what the outputs are not: the separate-report procedures are reported as answers to a set of mainly closed questions and "do not result in an opinion" under any auditing or assurance standard (pt 145). A clean answer set is not an audit opinion that your remediation was right — and the FAQ's own logic points the same way, since every one of these checks tests a policy you were supposed to have written before the error.

The FAQ, point by point — what each clarification actually settles

The FAQ is short (twelve questions across five chapters) but each one closes a real operational gap. This is the part the day-one law-firm alerts do not cover, because most of it did not exist until 24 December 2024.

What this means at the desk

Read the circular and the FAQ as one document, because the FAQ is where the money is. The circular gives you the thresholds and the form; the FAQ tells you that a lower internal threshold binds you, that "nobody was paid" still means "file", that a settlement-mismatch deposit breach is your fault not the market's, and that de minimis above EUR 25 will be questioned. Two filters decide whether any of the machinery runs at all, and both are yours to justify rather than assert: the tolerance threshold for NAV errors, and the active/passive call for investment-rule breaches — the second being the only one that removes the CSSF filing entirely (pt 55), which is exactly why the FAQ spends a third of its questions narrowing it. Neither filter reaches most of Chapter 6: the threshold is disapplied outright for a fee overcharge (pt 104), for a cut-off failure (pt 109), and it never appears in the allocation-error section at all — which makes "below our threshold" the wrong first question for three of the four other error types. The costliest single line remains pt 133 — correction costs cannot touch the fund — so the cheapest remediation is the control you evidenced before the error, and the de minimis you documented before you needed it.

The gotcha that catches people twice: the 4–8 weeks is a window for a complete filing, and the pre-notification is not a deadline extension you grant yourself. It is a request the CSSF can refuse — and a refusal does not give the time back, it sends you to a fresh complete notification with the clock already spent. Treat "we'll pre-notify" as an escalation to be justified in writing on day one, not a fallback you reach for in week seven.

Nothing in the circular has moved since it took effect — but the texts it points at have. Point 49's risk-spreading reference went with Circular 07/309 in December 2025, and Circular 22/811, the source of the administrator's error-handling duties in pt 23, was amended the same week to point at 24/856 in place of the repealed 02/77. The last of IML 91/75 went the same way in May 2026, taking the circular-level statement that UCITS investment limits apply to net assets with it. 24/856 itself is unreissued in all three cases, so the reader has to reconcile the cross-references; that is what the changelog below is for. One cross-reference moved the other way, early and quietly: the 21/790 separate-report procedures were rewritten for the circular in February 2025, exactly as pt 165 said they would be — the only text in the chain that changed because of 24/856 rather than around it. What has otherwise changed is the temperature. The CSSF's 2026 priorities for supervising the investment fund sector (31 March 2026) put valuation near the top under "asset valuation risk" — on-site controls on IFMs' valuation organisation, thematic sample-based reviews of open-ended private-assets funds including continuation funds — and add that "the correct implementation of Circular CSSF 24/856 concerning NAV calculation errors, instances of non-compliance with the investment rules and other errors at UCI level will be monitored by the CSSF". The practical consequence is a change of audience, not of rule: the internal policy 24/856 keeps deferring to — the documented threshold analysis under pt 35(d), the impact method fixed from launch under pt 72, the use and level of de minimis under pts 124–127 — is now a document an inspection may ask to see, rather than one only a notified error puts in front of the regulator. If it has never been written down, the first reader of it should not be a supervisor.

For how the same NAV error is treated in Ireland, the UK, Jersey and Guernsey — where the thresholds are lower, absent, or left to the depositary — see the NAV error correction matrix.