EU Clarifies CBAM Reporting for 2026 Importers

The Analysis

Brussels put the clarification out on 6 February 2026, and the first thing to establish is what kind of document it is. The European Commission issued guidance on Carbon Border Adjustment Mechanism reporting for importers in the 2026 period. Strictly speaking, guidance is not a legislative act. It does not create a duty that did not exist the day before, and it does not remove one. What it does is set out how the administration intends to read the duties that are already written into the underlying legislation. That distinction matters more than the headline, because a great many people will read reduce the administrative burden as relax the obligation. Those are not the same sentence, and confusing them is how a compliance plan ends up built on sand.

The date that should be on your wall is not 2026. It is 2027. The whole purpose of the clarification is to smooth the passage into the year in which the mechanism begins to bite financially. 2026 is a year in which you report; 2027 is a year in which you pay. So the reports generated in 2026 are not an administrative exercise that closes when you submit them. They are the evidentiary foundation for a cost that has not been invoiced yet. If the 2026 data is thin, the 2027 cost does not stay thin. It gets replaced by something worse: a value assigned to you by default rather than measured at the installation that actually made the goods.

The Commission's own framing is that the guidance lowers administrative burden while leaving the decarbonisation path intact. I read that as two statements that have to be held together. The administrative half is genuinely being worked on, and that is worth real money to anyone who has to produce these files. The decarbonisation half is not being traded away. If your 2026 plan rests on the assumption that the second half of that sentence will quietly erode under industry pressure, you are betting against the stated position of the legislator, and I would not want that bet sitting on my books.

It helps to be precise about where this document sits in the hierarchy. At the top is the regulation that establishes the mechanism itself. Beneath it sit the implementing and delegated acts that fill in the reporting mechanics, the calculation rules and the verification requirements. Below those sit guidance documents, frequently asked questions and administrative notices. Only the first two layers bind you. The third binds the administration, in the sense that the authority is expected to apply its own published reading consistently and you can normally rely on it being applied that way. But a guidance document is not something you can enforce against the authority if a later reading differs, and it is certainly not something that expands your obligations beyond what the regulation says.

I am deliberately not handing you article numbers here. Citation discipline matters in this area more than people think. If you are going to rely on a provision in a filing, a contract or a board paper, pull the number yourself from the Official Journal text in force on the day you rely on it. A number I give you from memory is worth less than no number at all, because it looks authoritative and may be one amendment out of date.

Now the applicable precondition, which is where most companies outside the EU get this wrong. The reporting obligation attaches to the importer in the member state where the goods are released. Where a customs representative is used and is acting in the indirect capacity, the obligation can sit with the representative rather than with the importer of record. The precise allocation is a matter of national implementation and of the representative's mandate, so read the actual appointment before assuming either way. What this means for a manufacturer outside the EU is uncomfortable and simple: you are not the addressee of the rule, and you are the only party who can produce most of the data the rule demands. The legal duty and the practical dependency are resting on different shoulders. That gap is not a legal problem. It is a contracting problem, and it is exactly the thing I would be working on during 2026 while the money is not yet moving.

Seen from the producer side, your position is stronger than you probably assume. The EU importer needs your installation-level data, because without it he falls back to default values, and default values in a mechanism of this kind are rarely better looking than measured ones. That is not a reason to hold anyone up for money. It is a reason to insist that the arrangement is written down: on what basis the data is provided, how often it is refreshed, whether a change of production installation triggers notice, whether an accredited verifier is allowed on site, and how responsibility is allocated if the numbers turn out to be wrong.

Those five points cost very little to settle now. Settling them in 2027, when the number is actually pricing a certificate obligation, costs a different order of magnitude. There is one more thing to understand about handing data over. Once your measured figures go to the other side, they become your own external statement too, and they have to be the same figure as the one in the importer's report and the one in the third-party verification statement. Three versions of the same number is not a documentation problem. It is a credibility problem, and the doubt it creates does not stay with the importer.

What is reported is embedded emissions. That is the greenhouse gas released in producing the good, split between direct emissions from the installation and indirect emissions from the electricity it consumes, plus the emissions carried in through certain precursor inputs. Two routes exist for arriving at a figure. One is actual values, measured at the installation and verified. The other is default values, published by the authority and applied when verified actual data is not available. There is a strong temptation to treat these as two options of equal standing and to pick whichever looks cheaper this quarter.

Don't rush to that conclusion. Default values are a fallback, and in most designs of this type they are set deliberately conservative so that the fallback is not more attractive than doing the work. The applicable precondition for using actual values is not that we have numbers. It is that we have numbers originating from a specific installation, computed under the approved methodology, and verified by an accredited verifier. Fall short on any one of those three and you are back on the default.

On evidence and burden of proof, the rule is short and it is not flattering: the party who wants to pay less carries the burden of showing why. That is the normal shape of these mechanisms, and it has a practical consequence that surprises people. When you claim actual embedded emissions, you are making an assertion about a factory you do not own, in a jurisdiction the authority cannot inspect at will, covering a production period that has already closed. The authority is not required to take that on trust.

So the file has to be built so that it can be re-performed by someone else. Production quantities per installation and the production route. Direct emissions and the activity data behind them. Electricity consumed, split by source where the source affects the factor. Precursor inputs, with the supplier, the quantity and the value taken for each. The emission factors used and the published source for each, with the version and the date you took it. The calculation file itself, in the form it was actually run rather than the form described in a memo.

The verification statement, the verifier's identity and its accreditation reference. And the gap log: where you did not have data, what you substituted, and who approved the substitution. That last item is the one people skip and the one that gets tested. If a third party cannot take your file and arrive at your number, your number is not evidence yet. It is a claim.

What a verification statement does and does not do deserves its own paragraph, because here there is an exception worth stating. A verification statement evidences that a verifier performed agreed procedures to an agreed level of assurance. It gives reasonable assurance about the data. It does not transfer responsibility for the data to the verifier, and it does not absorb the consequences if the figure is later challenged. Two things are worth checking on any verifier. Independence, meaning no conflict of interest with the installation. And scope, meaning whether the accreditation actually covers your sector and this class of emissions. A verification statement from a body whose accreditation does not reach your sector is worse than no statement, because it looks like evidence and does not survive being asked a second question.

On exceptions, I want to be careful. There has been provision in this area for very small consignments and for relief where the amounts at stake are trivial. Whether your traffic falls inside that relief is a question about the text in force at the moment of import, not about the text that somebody in your organisation happened to read last year. Two traps sit here. One is assuming the relief still exists in the form you remember. The other is more common and more damaging: assuming that because a consignment is relieved of the carbon obligation, the goods are outside the mechanism's scope altogether. Scope and relief are different layers. A good can be in scope and still relieved. A good can be out of scope and never relevant at all. Conflating the two produces the worst category of error, which is the one discovered three years later, on a different shipment, by a different auditor.

On timing. The rhythm is quarterly in design, and the year is cut into reporting periods that each close on their own schedule. I am going to resist giving you day-level deadlines here, because the concrete submission windows sit in the implementing rules and have been adjusted more than once, and a deadline repeated from memory is precisely the kind of detail that turns into a penalty. Read it off the official text for the period you are actually in. What I will say about time is structural, and it is the part that hurts. The reporting period and the payment period are not the same period, and the data you generate in the earlier one is what prices the later one. A late or sloppy 2026 quarter is therefore not a compliance blot that fades with time. It is a permanent input to a 2027 cost, and it will be sitting there when the certificates have to be bought.

On records, and I want to draw the line clearly. I am not going to tell you how to complete a return. That is not my lane and I do not give filing instructions. What I will describe is what an evidence file has to contain so that if the question is asked in 2028 about a 2026 quarter, you can answer it without reconstructing the past. Retain, per installation and per reporting period: production quantities and the production route; direct emissions and the activity data behind them; electricity consumed, with a split by source where the source changes the factor; precursor inputs with supplier, quantity and the value used; the emission factors and their published source, with version and capture date; the calculation file as run; the verification statement with the verifier's identity and accreditation reference; and the gap log described above.

On retention period, do not assume something short. Authorities can generally reopen a period years afterwards, and the national rules on how far back they can reach differ from one member state to another. Keep the file for the longest plausible reach-back across the member states you import into, not for your own accounting convenience.

A calculation, with every assumption stated up front. Assume an importer brings 1,200 tonnes a year of a covered good into one member state. Assume the installation's verified actual embedded emissions work out to 1.8 tonnes of CO2 equivalent per tonne of product. Assume the default value the authority would apply in the absence of verified data sits 15 per cent above that, at 2.07. Assume a carbon price of 80 euro per tonne of CO2 equivalent. Assume, finally, that obtaining verified actual data costs 20,000 euro a year, split between 12,000 of internal data work and 8,000 of verification.

On actual values the year's exposure is 1,200 times 1.8, which is 2,160 tonnes of CO2 equivalent. On defaults it is 1,200 times 2.07, which is 2,484. The gap is 324 tonnes, and at 80 euro a tonne that is 25,920 euro. Against a cost of 20,000 euro, the work pays for itself with 5,920 euro to spare. Now invert it, because the break-even is the number that actually travels. Divide the 20,000 euro cost by the annual saving per tonne, which is 0.27 tonnes times 80 euro, or 21.60 euro. The break-even volume is roughly 926 tonnes a year, about 232 tonnes a quarter.

Below that, on these assumptions, the verified route does not pay. Above it, it does. And note what the arithmetic is sensitive to. At a carbon price of 50 euro the saving per tonne falls to 13.50 euro and the break-even volume rises to roughly 1,481 tonnes a year. The answer is not fixed; it moves with the price. That is the reason the data programme and the price assumption belong in the same review, quarterly, rather than in two separate reviews that never meet.

What I would actually do, on a quarterly and annual scale. By the time the first quarter of 2026 closes, I want the importer-side ownership question settled in writing: which legal entity is the addressee, whether an indirect representative stands in its place, and which national authority is competent. That question has an answer in law, and it should not be answered by whoever happens to be copied on the email. By the close of the second quarter, I want a supplier data map: for each covered product line, the installation, whether verified actual data exists for it, and if not what the fallback is and who signed it off. The target I would set is verified actual data covering at least 80 per cent of embedded emissions by value, with every remaining line carrying a named owner and an approved reason for substitution.

Before the fourth quarter opens, I want one full reporting period run end to end on the live system rather than in a spreadsheet, because the first period you run on the live system is the period in which you discover what the system does not do. Before the year closes, I want the gap log reviewed line by line against the verification statements, because a substitution made in February becomes very hard to defend in December.

On alternatives, and I want to be careful here because I do not give commercial advice. Whether you move volume, change supplier or re-time purchases is your call, and I am not going to make it for you. What I can tell you is the legal consequence of each, so that the commercial decision is taken on the right facts. Changing a supplier does not reset your reporting history; the 2026 periods already closed stay closed and stay attributed to the entity that imported. Pulling volume into 2026 to get ahead of the 2027 charge shifts the year in which the cost is recognised, and whether that helps depends on the price at which certificates are actually sold in each year, which nobody can tell you today. Substituting a default value for missing data is not a neutral act either: it produces a number that is on the record, attributed to you, and available to be compared against the verified figures you file later.

The traps, then. Treating guidance as if it were law, in either direction, neither as a shield nor as a sword. Assuming that the practice of one member state's authority is the practice of all of them, when competence and national implementation differ and so does the practical tolerance for a thin file. Relying on a supplier's self-declaration with no verification statement behind it, when a self-declaration is an assertion by an interested party and is the weakest form of evidence there is. Assuming that the small-consignment thresholds used in customs valuation have anything to do with the carbon obligation, when they are different instruments written for different purposes and get conflated far more often than I would like to admit. And the deepest one: assuming that because 2026 is only a reporting year, a weak 2026 report is harmless. The report is the input. An invoice that arrives late does not make the input cheaper.

One last thing and then I will stop. The clarification was published on 6 February 2026 by the European Commission. Its stated purpose is to smooth the transition. Its stated constraint is that the decarbonisation path stays intact. That is the whole of what I am prepared to assert about it from the public record. Everything else I have said here is about how mechanisms of this general shape allocate proof and consequence, and about the arithmetic you can run on your own figures. If you need a provision, go and read the provision. My value to you is not that I remember the article numbers. It is that I know which ones you should be looking for, and what happens when the file cannot be produced.

↑ Back to the key points

— By Dr. Ingrid Voss