EU sets October deadline as China rejects 'market-closure' trade threats

The Analysis

Start with what actually carries a date. On 2 and 3 September 2026 the European Commissioner for Trade, Maros Sefcovic, warned that the European Union would deploy all trade defense tools if negotiations with China failed, and attached an October deadline to that warning. On 3 September, at a regular press briefing, MOFCOM spokesperson Huang Ling answered: consultations at every level must proceed on equal footing, neither side should make unilateral demands or dictate terms, and market-closure threats should not be used as bargaining chips. Those two statements are what everyone is working from. Neither of them is a legal act. I want to be precise about that before anything else, because the distance between a political deadline and a legal one is exactly where companies spend money they did not need to spend.

What does carry a legal date in this same news cycle is something else entirely. France's anti-fast-fashion law took effect on 1 September 2026. It imposes a tiered per-item environmental surcharge on products classed as ultra-fast fashion: EUR 0.25 to EUR 6 per item at present, rising to a maximum of EUR 20 per item by 2030, and capped at 50 per cent of the product's sale price. That is a member state statute. It is live. It has a rate you can put in a spreadsheet today. If you are looking for the thing in this story that actually changed your legal position this week, it is that one, and it is not the October deadline.

One observation of my own, having read a fair number of exchanges like this one. The temptation is always to treat the loudest sentence as the operative sentence. In trade law the operative sentence is almost always the one in a published notice, and it tends to arrive later, more quietly, and with a product definition attached that is either narrower or wider than anyone guessed. So the useful work in the quarter ahead is not deciding whether Brussels is serious. Seriousness is not a legal category. The useful work is making sure that if and when a notice does publish, the numbers you would need are already in a form you can hand over without a scramble.

Strictly speaking, a commissioner's warning creates no obligation for anyone outside the institutions he speaks for, and it creates no deadline that any court or any customs authority will recognise. I am not saying it is empty. A statement of that kind from that office tells you something about what is being prepared inside the Commission, and it would be foolish to ignore it. I am saying you should not build a compliance calendar on it. A trade defense measure comes into existence through a procedure: a complaint carrying the required industry support, an assessment of whether the complaint is admissible, a published notice of initiation that identifies the product, the period of investigation and the injury period, questionnaires, verification, and only then a provisional or definitive determination. Each of those steps has preconditions, and the preconditions are what you should be reading rather than the press line. If no notice has published in your product area, then as of today there is no measure applicable to your goods, whatever the deadline rhetoric says.

Where the October date does matter is narrower than the headlines suggest, and it matters procedurally. The reporting says Commissioner Sefcovic is scheduled for a video conference with Chinese counterparts in mid-September and an in-person visit to China in early October, with a ministerial meeting also set for early October. Those are negotiation events, not decision events. A trade defense measure is not adopted at a bilateral meeting; it is adopted through the Union's own internal procedure, on the record, with the reasoning published. What a negotiation deadline does is tell you when the political decision to start that procedure is likely to be taken. That is worth putting on your calendar, but it belongs on a different line from the one marked legal effect.

Now the applicable preconditions, in general terms, because this is where I see the most confident error. For an anti-dumping measure the elements are conventionally these: exports at less than normal value, material injury to the domestic industry, and a causal link between the two, with the injury assessment covering volume, price effects and the state of the industry. For a countervailing measure the elements are a subsidy that is specific, plus injury and causation. For a safeguard the shape is different again, and it is generally directed at import volumes rather than at unfair pricing. I am describing the architecture, not reciting provisions.

I am deliberately not giving you article numbers, and I would be doing you a disservice if I did: the exact tests, the de minimis thresholds, the standing requirement for complainants, the definitions of the like product and of the Union industry, all of these sit in texts that have been amended over time and that you should read in the version in force on the day it matters. Pull them yourself. A citation I give you from memory is worth less than no citation, because it looks authoritative and may be one amendment out of date.

The timing question, and there is a genuine exception here, so let me be careful. The ordinary position in trade remedy work is that a measure applies to goods entered after it enters into force, and that goods entered before that point are not caught. Do not let that reassure you too much. Several trade remedy systems provide for registration of imports during an investigation, and for the retroactive application of duties to registered imports once certain conditions are met, within a defined window before the measure is imposed. Whether that mechanism exists in the instrument that would apply to you, what triggers it, and how long the reach-back window is, are all questions answered by the specific text and by the notice that publishes in your case.

I am not going to assert any of them. What I am telling you is that the assumption that you have a free run until the day a measure is published is not safe in every system, and the way to find out is to read the registration and retroactivity language in the initiating notice the moment one appears covering your product.

There is a second timing point that people miss, and it runs in the other direction. A measure, once imposed, normally stays for years rather than months, and it is subject to expiry review and to interim review. That is where my time horizon sits. If you are an exporter whose goods could fall within a definition, the decision you make in the next quarter about what records you keep will still be governing your position at an expiry review five years out, and possibly at an interim review before that. And if a measure is imposed and you want to contest it, the route is an action before the Union courts, which is measured in years, not quarters. Anyone planning on a trade remedy question resolving inside a single financial year is planning on something the procedure does not offer.

Let me separate the French statute out, because mixing it into the EU story is the most common error I expect to see this month. The French law is national legislation of a member state. It took effect on 1 September. Its mechanism is an environmental surcharge per item on products falling within a national classification of ultra-fast fashion, and the reporting identifies SHEIN, Temu and AliExpress among the platforms it targets. It is not an EU trade defense instrument, it was not adopted through the Union procedure, and it does not depend on any finding of dumping, subsidy or injury. It applies because a national legislature passed it and it commenced.

If you contest it, you contest it under French law and, at the state-to-state level, under the WTO covered agreements. Whether it is WTO-consistent is a contested question, not a settled one. MOFCOM has said the law may violate the WTO non-discrimination principle and has described it as discriminatory; that is a stated position by one government about another's measure. It is not a ruling. No panel has decided it, and I am not going to tell you it is illegal, because I do not know that and neither does anyone else yet.

On the Chinese side, the reported position is that if France persists, China will take necessary measures to safeguard the legitimate rights and interests of Chinese enterprises, and that all consequences shall be borne by the French side. What form any such measure might take, whether it would be directed at French agricultural, luxury or cosmetics exports, or take some other shape entirely, is not something I can tell you and not something you should plan around as though it were decided. A retaliation decision is a decision by a competent authority taken through its own procedure and published. Until it is published it is a possibility, and building a procurement plan on a possibility is how a company ends up hedging against a risk that never materialised while ignoring the one that did.

Also in the reporting, and worth one paragraph because I expect it to be over-read: on 3 September China and Egypt signed two memoranda of understanding, one on strengthening supply chain economic cooperation and one on developing the Suez Economic and Trade Cooperation Zone, and China signed a joint statement with Kyrgyzstan on service trade and investment agreement negotiations plus an e-commerce cooperation memorandum. A memorandum of understanding is, in the ordinary case, a document recording an intention to cooperate. Whether it creates enforceable rights and obligations depends entirely on its terms and on the intent expressed in it, and many do not. Read the text before you treat any of these as a market access commitment or as a basis for a sourcing decision. I am not saying they are without value. I am saying the legal weight of an MoU is a question about the document, not a question about the photograph of the signing.

Evidence and burden of proof now, which is the part of this that actually decides outcomes and the part companies leave until it is too late. The structure in a trade remedy investigation is not the structure of a commercial dispute between two companies. At initiation the complainant carries the burden of showing enough to justify opening an investigation. Once it is open, the investigating authority gathers the record, and the principal route by which your own facts get onto that record is the questionnaire response and the verification visit. If you do not reply, reply late, or reply with figures that cannot be reconciled to your own books, most systems permit the authority to base its findings on the facts otherwise available, and in practice that means it may use the complainant's figures or any other reasonable source.

The consequence is not neutral: a non-response tends to produce a worse result than a response containing bad news. That is the single most useful sentence in this whole piece. Refusing to hand over unhelpful data does not make the data go away. It hands the calculation to somebody else.

So what should be kept, and in what shape. I will describe the file rather than prescribe a process, because the process is not mine to advise on. The sales side, transaction by transaction: invoice number, date, customer, whether the customer is related, quantity, unit price, currency, the invoice currency and the conversion method applied, the delivery term, and every rebate, commission, credit cost and after-sales provision attached to the sale. The comparison side: the domestic price list for the same or like product over the same period, with a documented list of the adjustments made to render it comparable, freight, insurance, packing, credit, warranty, level of trade, and the basis for each adjustment.

The cost side: a bill of materials and a consumption record per product, showing input quantities and unit costs, plus the allocation method for labour, energy and overhead, and a reconciliation showing that the allocation method used in the questionnaire is the same one used in the statutory accounts. The subsidy side, which is what catches people: a complete inventory of every grant, tax preference, preferential loan, below-market input, land-use benefit or export-related incentive received by the entity and by any upstream supplier whose inputs are material, with the programme name, the granting authority, the legal basis cited at the time and the amount by year.

The volume side: capacity, production, capacity utilisation, opening and closing stock, and the order book, by product and by month. The governance side: a data dictionary stating what each figure means and where it comes from, a change log recording every restatement, and a named person accountable for each table. If a verifier takes your file and cannot recompute your number, the number is not evidence. It is a claim.

Three calculations, with the assumptions written out, because vague numbers are worse than no numbers. The macro one first. China-EU trade over January to July 2026 is reported at RMB 3.67 trillion, up 9.5 per cent year on year. Assume the monthly average of those seven months continues for the remaining five. Then the full-year figure would be 3.67 divided by 7, multiplied by 12, which is about RMB 6.29 trillion. Assume instead that the year-on-year growth rate of 9.5 per cent is the better guide, and that the prior-year comparable was 3.67 divided by 1.095, about RMB 3.35 trillion, implying an increase of roughly RMB 320 billion in the first seven months alone. Both are linear extrapolations and I am labelling them as such. Neither is a forecast, and neither tells you anything about your own exposure, which depends on your product and your share of it.

The unit economics one, using the French surcharge because it is the one with a live rate. Assume a garment with a sale price of EUR 20. Assume the surcharge applies at the current top of the reported band, EUR 6 per item, which is 30 per cent of the sale price. Assume landed cost of EUR 6 and platform plus fulfilment cost of EUR 5, leaving gross margin of EUR 9, or 45 per cent. With the surcharge, margin falls to EUR 3, a reduction of 67 per cent. Now assume the surcharge reaches the reported 2030 ceiling of EUR 20 per item, and assume the 50 per cent cap bites, so the actual charge is EUR 10.

On the same cost structure, margin becomes EUR 20 minus 6 minus 5 minus 10, which is minus EUR 1. The product loses money. That is arithmetic on stated assumptions, not a prediction about 2030 and not advice about what to charge. It does tell you what kind of cost change has to be absorbed somewhere in the chain, and it tells you the answer is not a small one.

The third one, on duty sensitivity, because it is the calculation I want on a page before any notice publishes. Assume a company exporting RMB 50 million a year to the EU. Assume a definitive duty of 20 per cent arrives and assume, for the sake of the arithmetic, that none of it can be passed through to the customer. Annual added cost is RMB 10 million. Assume the EU business runs at a net margin of 8 per cent, which is RMB 4 million. The duty wipes out the profit and leaves a shortfall of RMB 6 million. Now assume variable cost is 70 per cent of export value, RMB 35 million, and ask what cost reduction would close the gap: 10 divided by 35 is 28.6 per cent.

Assume that the maximum variable cost reduction realistically achievable inside twelve months is 8 per cent, which is RMB 2.8 million. That leaves RMB 7.2 million still to find through some combination of price, volume and mix. I am not telling you which combination. I am pointing out that on these assumptions no single lever is large enough, and that discovering this in the month a measure is imposed is nearly a year later than is useful.

Alternatives, and what each costs. The first alternative is to do nothing until a notice publishes, on the basis that nothing is yet binding. It costs you the cheapest window you will get: the period before initiation is the only period in which you can assemble historical records while the people who made them are still in the building and the systems they were made in are still running. The second is to prepare fully for the worst case across the whole product range. It costs you a large amount of work on products that may never be covered, and I have seen companies exhaust their compliance capacity on a scare and then be unable to respond properly to the notice that actually arrived.

The third, and the one I would take, is a triage: rank product lines by exposure to the definition you think likely, and put the file together for the top of that list now while keeping the rest at a state where it can be assembled quickly. It is not as clean as either extreme, and it is the only one that survives contact with a real calendar.

The traps. Treating a political deadline as a legal one and reorganising supply chains against a date that no authority will enforce. Treating a member state statute as an EU measure, or the reverse, and thereby looking for the wrong remedy and the wrong forum. Writing a contract clause that shifts a new duty to the other party without identifying the duty, the taxable event or the person liable, which produces a clause nobody can perform. Running three sets of numbers, one for customs, one for the statutory accounts and one for the questionnaire, on the assumption that nobody will line them up; somebody will.

Deciding, on the advice of somebody who is not a lawyer, that a measure is obviously inconsistent with WTO rules and therefore safe to ignore, when the only thing that makes a measure unlawful is a ruling that says so. Assuming that because talks are described as candid and constructive, no measure is coming; the two facts are independent. Assuming the product scope will track your HS codes, when scope is set by a definition written into a notice and often cuts differently. And, on the Chinese side, assuming a retaliatory measure has been decided because a spokesperson said necessary measures would be taken.

On timing, which is where I would leave you. This quarter, do the triage and the file skeleton: name the product lines, name the accountable person, and get the data dictionary written so that when a questionnaire lands you are filling in blanks rather than deciding what the blanks mean. Over the following year, get the file to a state where a verifier can recompute your figures from source and where the cost allocation used in the response is demonstrably the same one used in the accounts. Beyond that, put the review dates on a multi-year calendar, because the measure you are worried about will outlive the news cycle that produced it by several years, and because the record you build now is the record that will be tested at an expiry review long after everyone has forgotten what was said in September.

I am not going to tell you to reshore, to reprice, or to move volume. Those are commercial judgements and they are not mine, and I have no interest in dressing one up as legal analysis. What I will say is that the only part of this story with a rate, a commencement date and a mechanism you can put in a model is the French statute, and it has already commenced. Everything else is procedure, and procedure is something you can prepare for even when you cannot predict it.

↑ Back to the key points

— By Dr. Ingrid Voss