Cross-Border: Company Formation as a European Use Case

So far we have tested digital twins of legislation on Austrian levies: on the tourism levy of one federal state and on the federally regulated municipal tax. Both cases stop at the national border. Company formation does not: anyone founding a company in a member state of the European Union today does so under national law that transposes European directives, and with evidence drawn from the registers of other states. This article explains why company formation is the right next use case, what role registers and ontology play in it, and what that means for our research agenda.

Formation: a business situation with many sources of law

In "Der selbstfahrende Staat" (Springer Gabler 2024) we organized the state's services for companies along 16 business situations, and formation comes first. In the analog state it meant a notarial appointment, a declaration of formation drafted by a lawyer, the specification of industry and purpose, and, where required, a trade license, before the company could be entered in the commercial register. The digital state records master data electronically and verifies identity through ID Austria or Germany's BundID, but still registers separately with the commercial register, the tax office, and social insurance. The automated state offers a standard formation process in real time in which all data reach the competent authorities immediately. The self-driving state advises: from structured information provided by founders it determines suitable legal forms, names, and trades, registers automatically upon consent, and triggers the follow-up processes, from tax registration to funding applications.

What makes this progression interesting for our research is the number of legal sources that converge in a single business situation. The Austrian Limited Liability Companies Act (GmbHG) governs legal form, share capital, and formation; the Commercial Register Act governs registration; the Trade Regulation Act (Gewerbeordnung) governs the license and the proof of qualification; tax law governs fiscal registration; social insurance law governs enrollment. Above all of this sits Union law: Directive (EU) 2019/1151 obliges member states to enable the fully online formation of certain types of limited liability company, in Austria the GmbH, and to complete it within five working days where only natural persons found the company using the prescribed templates, and within ten working days otherwise. A formation procedure is therefore not one law but a bundle of laws, and bundles like this are exactly what we want to translate into twins next.

Not everything in that bundle is computable. The first feasibility condition from our interview study, described as the first of the four core categories in the article Explainable automated decisions from the digital twin of legislation, requires automating only deterministic and parameterizable sub-decisions. For formation, that line can be drawn cleanly:

• Deterministic. Whether a single-member limited liability company may use the simplified formation under Section 9a GmbHG without a notarial deed; whether share capital and cash contribution meet the minimum values; which fees apply and whether the exemption under the New Business Promotion Act (Neugründungs-Förderungsgesetz) applies; whether an activity is a free or a regulated trade; which deadlines are running.
• Discretion. Whether a company name is distinctive and not misleading; whether an individual qualification under Section 19 of the Trade Regulation Act exists; whether a qualification acquired abroad is equivalent where no automatic recognition applies.

The twin covers the first list and hands the second back to people. That is not a limitation of the use case but its precondition.

What the single market changes

For the tourism levy we showed that a centrally provided, executable version of a law reduces the collective implementation burden roughly 35-fold compared with decentralized re-coding; the derivation is in the article What implementing legislation really costs. That case concerned one state law and around 29,000 businesses liable for the levy. The single market multiplies the problem: a directive is transposed in 27 member states, and each transposition act is in turn translated into code by register authorities, notaries, chambers, software vendors, and companies. The same normative idea, for example that a limited liability company can be formed entirely online, ends up in hundreds of implementations that know nothing of one another.

Over the past years the Union has done a great deal to ensure that at least the evidence travels. The single digital gateway under Regulation (EU) 2018/1724 connects the procedures of the member states and, since December 2023, obliges them to exchange evidence across borders under the once-only principle. The Business Registers Interconnection System (BRIS) makes commercial register data retrievable across Europe. Directive (EU) 2025/25, published in the Official Journal at the beginning of this year, introduces an EU Company Certificate and a digital EU power of attorney so that a company registered in one member state can prove its existence and its representation in any other. Regulation (EU) 2024/1183 provides for a European digital identity wallet for citizens and companies. And Regulation (EU) 2024/903, the Interoperable Europe Act, creates the framework in which public bodies connect their services across borders.

All of these legal acts govern the transport of data. None governs the machine-readable version of the rules applied to that data. Anyone who today examines a formation involving parties from another member state receives the evidence through an interface and the provision as prose.

The evidence is already traveling across Europe. The rules that evaluate it are not yet on the road.

Register logic: the twin of the company meets the twin of the law

A twin of legislation computes only as well as its inputs. In our interview study with nine experts from EU legislative bodies, a national parliament, academia, notarial practice, and an ERP vendor for the public sector, interoperability emerged as a pillar of its own: standardized data exchange, unique identifiers for persons, organizations, and assets, and connected registers, so that the twin obtains the inputs it requires. "E-invoicing and integrated registers are enablers," one interviewee said. For formation, the inputs are clearly named: identity and civil status of the founders, the criminal records register (Austria's Strafregister or Germany's Bundeszentralregister) for reliability in certain trades, the commercial register (Firmenbuch or Handelsregister) for participating companies, proof of qualifications for regulated activities.

In the book we described the digital twin of the company for this purpose: it is created at formation, exists until closure, and keeps business data, legal documents, and financial reports available to the administration; the existing registers store its master data. The twin of legislation is its counterpart. One supplies the facts, the other the rules, and the decision arises where the two meet: because a specific provision of the transposition act applies and the inputs from a specific register read as they do, the result is what it is. This explanation is not a justification written after the fact but the log of the computation, and it works across borders as soon as both sides use the same identifiers.

The business situation of foreign trade in the book shows what this can look like. In the analog state, anyone wishing to offer services across Europe must notify the authorities with paper evidence of nationality, establishment, trade license, and insurance coverage. The automated state provides the evidence from a federation of states and initiates qualification steps only if the automated rule check finds a problem. That rule check is precisely a twin of legislation.

One ontology for 24 official languages

The second feasibility condition, semantic standardization and ontological alignment, becomes the main task in the single market. One expert in our study reminded us that even the term "child" is defined differently in tax law than in criminal law. Company law across the member states is no different: registered office, establishment, management, shareholder, and power of representation carry similar names and mean different things. The interviewees recommended three things that fit together for a European use case: anchoring definitions at the higher legislative level so that subordinate rules inherit them; language-neutral identifiers to align multilingual versions; persistent URIs under which the meaning of a term can be retrieved. Two European building blocks already exist: the IATE terminology database as a starting point for concepts, and the European Legislation Identifier (ELI), which makes every provision of a legal act unambiguously and versionably addressable across jurisdictions.

Applied to the four layers from the article From prose to model, a cross-border twin is distributed as follows:

This corresponds to what we defined in our layer model as the configuration layer: jurisdiction-specific parameters such as rates, thresholds, and lists are maintained as data, not as code. For the single market this means that the logic of a check is written once, the values come from 27 configurations, and each configuration points to its transposition act.

What the research agenda makes of it

We already named cross-jurisdictional applicability as an open research item at the ACM Symposium on Computer Science and Law in March 2025, and our interview study ends with the same recommendation: cross-border pilots, standardization of the ontology, governance models. Company formation meets every criterion we apply to a use case. It is deterministic in its core questions, it is already embedded in Union-law procedures, its inputs are available in registers, and it affects every company exactly once, but every one.

For a European pilot, five work packages follow:

1. Delimit. Identify the deterministic fragments of the directive and the transposition act, starting with eligibility for online formation, deadlines, and fees.
2. Anchor. Address every provision at Union and national level through ELI and explicitly model the transposition relationship between article and section.
3. Align. Build a core ontology of company law that connects to IATE and allows national extensions; our ongoing systematic literature review of legal ontologies provides the foundation.
4. Connect. Couple the twin's inputs to once-only evidence and register queries so that the explanation of a decision points to the norm and the data source at the same time.
5. Assign. Clarify who maintains and certifies the twin of a directive and who maintains the twin of its transposition, and place that responsibility with a competent public unit staffed with legal and technical expertise. Without this governance, the fourth condition of our study, adapted legislative and governance processes, remains unmet.

In its interoperability strategy for 2021 to 2027 the European Union has discussed publishing machine-readable regulations, and individual states such as France and Finland have experimented with coding tax and social benefit rules for internal use. A twin goes further because it turns the executable version into a public artifact that is traceable to the wording. I consider formation to be the case in which this can be shown for the first time in several legal systems at once, because thanks to once-only the evidence is already on its way and only the rules are missing.

The next step

The foundations of this article, the four layers, the feasibility conditions, and the burden model, can be found in the previous articles of this series and in our papers under Publications; the vision behind them is described on the page Digital twin of legislation. If you work on online company formation in a register authority, a notary's office, a chamber, or a ministry and would like to help shape a cross-border pilot, we invite you to talk to us.