Europe wrote America's way in

On 26 June 2026, South Korea's president Lee named two companies: Palantir and Helsing. Not two programmes, not two capabilities. Two private firms, held up as the benchmark to match.

Korea can afford the comparison. That same month Seoul announced a strategic investment vehicle, Korea Strategic Technology Partners, with a ceiling of 10 trillion won over five years, roughly $7.2 billion, aimed at sovereignty technologies, defence included but not exclusively. It also has the diplomatic frame: the European Union and the Republic of Korea signed a security and defence partnership on 4 November 2024.

What Korea does not have is a SAFE agreement.

Canada does. Negotiations closed on 1 December 2025, signature on 14 February 2026, formal conclusion by the Council on 15 June 2026. The third country route exists, it works, and it has produced a result. For Korea, so far, nothing.

The contrast counts less as a grievance than as a marker. It shows where access to the European defence market is actually settled. Not in the quality of the product, not in diplomatic proximity, not in the force of the story: in a text adopted in spring 2025 that the world discusses as a political signal and that almost nobody reads line by line.

This issue reads it.

SAFE sorts fundable equipment into two categories, and the line between them turns on something founders almost always underrate: design authority. It does not mean owning a patent, and it does not mean where the factory stands. It means control of the technical definition of the product: who can modify it, adapt it to a national requirement, integrate it into another system, keep it in service for twenty years without asking a third party for permission. In defence, that is the line between a buyer and a partner.

The first category is exempt from the requirement. Missiles sit in it. Five weeks before adoption, the body that represents Europe's aerospace and defence industry asked in writing for them to be moved into the second. The Council kept them in the first, with no published reasoning for the gap.

Thirteen months later, Anduril set up production of a cruise missile in Bydgoszcz, without giving up its design, in compliance with the text that the sector's umbrella body had tried to close on this exact point.

The regulation gates access to funding through two separate mechanisms, and the confusion between them accounts for a good share of the commentary published over the past year.

The first is a cost threshold. The adopted text caps the share that falls outside the eligible zone: no more than 35% of the estimated cost of the components of the final product. The base is the components. It is not the product.

The second is a control requirement. Article 16 sets out two lists of equipment categories. For the second list, access to funding requires European design authority. For the first, that requirement falls away. Missiles are on the first list. So are cyber and the protection of critical infrastructure. Command and intelligence systems, which the text calls C4ISTAR, are on the second.

The two mechanisms carry nothing like the same weight. A cost threshold can be managed: you relocate a component, you change supplier, you rebuild a bill of materials. Design authority cannot be managed. It is kept or it is handed over. A manufacturer who keeps it keeps the product; a manufacturer who transfers it becomes the producer of somebody else's object. Putting a piece of equipment in one category rather than the other therefore settles the nature of what a state will buy, and of what it will own ten years from now.

One figure has been circulating everywhere since spring 2025 all the same: 65%. It appears nowhere in the text in force. It comes from the Commission proposal of 19 March 2025, which required no less than 65% of the estimated cost of the final product. Between March and May the provision went through three changes at once. The direction reversed, the floor became a ceiling. The base narrowed, the product became the components. And the figure vanished. On 18 June 2025 a corrigendum in the Official Journal further repaired the French version of recital 19, which had said "lower" where the text meant the opposite.

Three reference analyses currently cite three different bases. Federico Santopinto reads a 65% components requirement; a widely circulated Ukrainian analysis reads 65% of the final product; Dariusz Czuchaj gives a third variant. None of them matches the text in force. These are three serious readings, by people who know the subject. That all three diverge says something about the text, not about their authors: it changed meaning in nine weeks, and its French version had to be repaired after publication.

Content rules are not a European invention. The Buy American Act of 1933 has tied US public procurement to a share of domestic production for close to a century, and that scheme reads first through its exceptions: public interest, nonavailability, unreasonable cost. What decides, in this kind of text, is never the headline threshold. It is the list of what escapes it.

Europe has written content rules before. It has just written them at a scale of 150 billion euros. It wrote the threshold. It also wrote the list.

On 22 April 2025, the AeroSpace and Defence Industries Association of Europe published its position on the instrument then under discussion. The document deserves to be read in full, because it does not say what a story of conflict would want it to say.

ASD welcomes the proposed regulation. It explicitly endorses the principle of two differentiated categories: the idea that European design authority can be required for some equipment and not for other equipment suits it. Then, introduced by a plain For instance, it asks for three headings to be moved into the second category: missiles, the protection of critical infrastructure, and cyber. And it offers a fallback of its own, should the request not go through: a prioritisation mechanism, or financial incentives, for products whose design authority is European.

The three elements count together. Keeping only the request would stage a confrontation that never took place. Industry is not attacking the text, it is backing it, it accepts the architecture, and it plans its way out in advance in case the adjustment is set aside. That is how an organisation behaves when it wants the instrument to exist.

Which leaves the request itself, and what it costs the party making it. Moving from category one to category two amounts to asking for an extra constraint on products its own members build. The association asks for no exclusivity, no national preference, no closed market. It asks that European public money spent on missiles go to missiles whose design is controlled in Europe. And it asks this on three specific headings, chosen, not on the whole list. An organisation that prepares its fallback at the same time knows it may not be followed.

On 27 May 2025 the Council adopted the text. The three headings stayed in category one. No published reasoning explains the gap between the request and the final text.

That last sentence will be read too fast, so it is worth fencing. The Council owed no line by line reply, and the absence of published reasons is neither an irregularity nor a concealment: a regulation adopted under emergency on an exceptional legal basis does not come with a public debate heading by heading. The finding is about what exists and what does not. There exists a written, public, dated request from the organisation that speaks for the sector in Brussels. There exists no public document explaining why it was not followed.

The chronology, for its part, protects the analysis. The three headings were already in category one in the Commission proposal of 19 March, a month before the ASD position. So the Council did not rule against European industry: it kept a line the Commission had already drawn. The nuance does not dissolve the fact, it hardens it. A classification proposed by default, contested publicly and in writing by the sector's umbrella body, then adopted unchanged and unexplained, stops being a classification by default. It becomes a classification confirmed.

Nobody has made this connection. Santopinto identifies the missile anomaly and treats it as an inconsistency of the compromise. Czuchaj describes the indeterminacy of the categories and how unevenly they are applied across countries. The IISS documents third country applications. None of the three links the document of 22 April to the vote of 27 May.

The oversight is not mysterious. A trade association position reads as a lobbying document: you take the request and you file it. A regulation reads as a final state: you analyse the mechanism. Nobody sets the two documents side by side on a timeline, because they belong to two different kinds of reading. Yet that is the one operation that changes the interpretation: it turns an oddity of drafting into a documented decision.

The anomaly is not a compromise gone wrong. It is a decision.

The first reflex, faced with a text like this, is to gauge how hard it would be to correct. The answer is surprising: not hard at all.

SAFE rests on Article 122 of the Treaty on the Functioning of the European Union. A Council regulation, no co-decision, no passage back and forth through Parliament. Revising it follows exactly the same route: a Commission proposal, the Council alone, by qualified majority. The Council adopted SAFE in nine weeks, from 19 March to 27 May 2025. It could reopen Article 16 at the same speed. The lock is not a legal one.

The asymmetry is the point. Standard commentary takes the speed of adoption as proof that Europe can finally move fast on defence. It proves the reverse just as well: a text written in nine weeks can be corrected in nine weeks. Fourteen months after adoption, no correction has been proposed on the categories. That silence owes nothing to procedural inertia: there is no procedure to overcome. Every month that passes renews the decision.

The European Parliament is attacking the regulation all the same. On 20 August 2025 it filed an action for annulment against the Council, case C-560/25, on the choice of legal basis. Two details of that action weigh more than the action itself. It does not concern the equipment categories. And it expressly asks that the effects of the regulation be maintained until it is replaced. The institution challenging the text before the Court is asking the Court to let the text keep working.

That leaves the member states. According to the work of Federico Santopinto, published by IRIS in June 2025, the negotiation set three lines against one another. Poland opposed the very concept of design authority, its industrial doctrine resting on technology transfer. Germany and the Netherlands were open to American licences produced in Europe. France defended European intellectual property. No primary institutional document formalises these positions: they come from research work, and Cellule 51 cites them as such.

Poland's consistency can be read without any need to judge it. Warsaw is the instrument's largest beneficiary, with 43.7 billion euros allocated and a first disbursement of 6.6 billion on 29 May 2026. Its doctrine aims at production lines on its own soil, industrial jobs and transferred skills, rather than at title over a set of drawings. A European design authority requirement would have shut the door on part of the equipment it wants assembled at home. It did not want that requirement, and it accounts for the largest volume of the instrument.

Nobody, in this configuration, gains by reopening the file. Not the buyer states, which got what they wanted. Not Parliament, which asks for the stability of the text it is challenging. Not the Commission, which would have to propose the revision of an instrument presented as its own success. Industry has already spoken, and it had planned its fallback.

The anomaly was flagged in writing. It stood.

On 6 July 2026, in Bydgoszcz, the Polish state group PGZ, its subsidiary WZL-2 and Anduril signed a cooperation agreement. Assembly first, production later. The product is the Barracuda-500M. The announcement came from defence minister Kosiniak-Kamysz.

Anduril presented that product family on 12 September 2024, under the name Barracuda-M Family of Cruise Missiles. Eight months before SAFE was adopted, twenty one months before the Polish agreement. The designation "missile" is native, commercial, and older than the regulation.

The reading runs downward, and only downward. The regulation puts missiles in category one. Category one is exempt from European design authority. The Barracuda-500M is a cruise missile by a designation that predates the text. It therefore enters through that category without giving up anything of its design, and no transfer of design authority appears in either party's announcement.

No intent is attributed to anyone. The easy story would be an American manufacturer finding a loophole in a badly drafted European text. The chronology rules it out: the product designation precedes the regulation by eight months, and nothing indicates that either was built around the other. What remains once that story is removed weighs more than the story did. The grid did not open for Anduril. It was written around products that already existed, with their names, their commercial classification and their technical files. The choice belongs to the Council.

One point of status, because it matters for what comes next. Both parties present SAFE compliance for the cooperation as an objective to be reached in phases, not as an established fact. No contracting authority has qualified the product to date. What is established is the category in the text and the designation of the product. The rest belongs to the calendar.

What the decision produces looks like this: Poland gets assembly lines and a transfer of industrial know how, in line with its doctrine; Anduril gets access to the Union's largest defence funding instrument, with no concession on its intellectual property; European industry gets the regulation it had welcomed, minus the adjustment it had asked for. Nobody was deceived. Everyone got what the text provided for, including those who would have preferred it to provide for something else.

There is a floor here that narrative analysis rarely reaches. In defence, the same actor writes the threat and signs the cheque, and the story decides which of the eligible players public procurement notices first. This issue adds the step below. Before it selects, the state draws the boundary. It writes the list of what can be funded, the category where the requirement applies, and the one where it falls away. The best story in the world does not move a product from one list to the other. No campaign, no funding round, no press coverage has ever reclassified a heading in an official journal. Narrative competition takes place inside a perimeter somebody drew beforehand, and that perimeter is read in an article of a regulation, not in a press release.

For a European founder, three questions, in this order.

Which category covers the product, and who qualifies it? The question comes before positioning, before the story, before the pitch. It settles what will have to be given up to reach the funding. The text provides no tie breaker for borderline cases, and Dariusz Czuchaj has documented that indeterminacy along with how unevenly it is applied from one country to the next. The most exposed case is software. The word "cyber" has been in category one since the first version of the text and has never been defined. The acronym C4ISTAR is absent from the Commission proposal: it turns up in category two during the negotiation. A command software package can plausibly fall under both. The grey zone is a product of the negotiation, not a slip of the pen.

How is the cap actually calculated? On the estimated cost of the components, not on the cost of the finished product. The difference is heavy: it changes the base, therefore the calculation, therefore the architecture of the supply chain. The bill of materials becomes a regulatory document. A founder who has built the case on the final product has the wrong denominator, in good company, since three published analyses make the same mistake in three different forms.

What does compliance guarantee? Nothing beyond eligibility. It opens the door, it does not win. Once inside the room, the mechanism from the first issue of this Wire takes over again, and it knows nothing about categories. Regulatory compliance has never selected anyone: it establishes who is allowed to compete, not who takes the contract. A founder who builds a strategy on eligibility has done half the work, and done it out of order. A category is established first and negotiated second.

For an investor the consequence is simpler still. A product's SAFE category is now a diligence item, on the same footing as the intellectual property structure. It states what the company will have to concede to reach European funding, or what it will not have to concede. Two companies at the same stage, on the same technology, do not carry the same risk if one falls in an exempt category and the other does not. And the follow up question is the tie breaker: on a product that can fall under both lists, value depends on a qualification that no text defines and that each national contracting authority will handle in its own way. A diligence note that covers that point in one line covers in one line what will decide access to the continent's largest defence funding instrument.

The bow and the lyre share the same string. So does the SAFE regulation: it requires European design on one side of its grid and waives it on the other. A text that does both is not contradicting itself, it is choosing. And that string was drawn by an identifiable hand, on a known date, after a public warning.

Europe gave itself an instrument of sovereignty and wrote into it, in as many words, the categories where sovereignty would not apply. The world reads that text as a model. It is first of all a price of entry, and that price was set in full knowledge.

What the text does not settle is still sitting there. Where the grid hesitates, in command software for example, another currency decides who gets in and who stays out, and that will be the subject of the next issue.

Polemos pater panton.