The ENISA favourable report has become one of the most important documents for a technical founder relocating to Spain, because it is the evidence that supports the innovative-entrepreneur route — one of the qualifying reasons that can open the Beckham Regime under Article 93 and the wider entrepreneur residence framework of the Startup Law. In the current wave of relocation, a large share of these founders describe their company as an "AI startup". That label is doing a lot of work, and the uncomfortable truth is that it does very little work in the assessment itself. The report tests whether a project is genuinely innovative and of special economic interest; "we use AI" is a starting sentence, not an answer.
This note is deliberately narrow. It is not a general explainer of the ENISA report and the Beckham Regime, nor the broader guide to the Beckham Regime for AI and software founders. Its single job is to explain what makes an AI project actually look innovative to an assessor — and, just as usefully, what makes one look like AI-washing. If you are building on top of a general model and hoping the sector label carries the file, this is the piece to read before you apply.
On this page
Why "AI startup" is not evidence Technical substance: what you actually built Defensibility and the wrapper problem Economic interest and scalability Team and capacity to execute Showing it, not asserting it How this connects to Beckham Frequently asked questions
"The founders who breeze through are never the ones who repeat 'AI-powered'. They are the ones who can show, in a page, exactly what they built, why it is hard to copy, and why it matters economically."
— Jacob Salama · International Tax lawyer, Ilustre Colegio de Abogados de Málaga (nº 11294)
Why "AI startup" is not evidence
The assessment behind the favourable report is looking for two things that travel together: innovation and special economic interest. Neither is proven by naming a technology. In a period where a very large number of new companies attach "AI" to their description, an assessor cannot treat the word as a signal of innovation, because it no longer distinguishes anything. What distinguishes a project is what sits beneath the label — the technical contribution, the defensibility, the market and the team. A file that leans on the sector name, with slides full of "AI-powered" and little underneath, is a weak file precisely because it mistakes vocabulary for substance.
This is not scepticism about AI; it is the opposite. Genuinely innovative AI companies exist in large numbers, and they tend to have no difficulty describing what is novel about them. The problem is the crowded middle: projects that use a general model to do something reasonable but not distinctive, and then present that as innovation. The report is designed to separate the two, so the founder's task is to demonstrate which side they are on, in concrete terms an assessor can evaluate.
The word "AI" no longer distinguishes anything. What the report evaluates is what sits underneath it — the technical contribution, the defensibility, the market and the team.
Technical substance: what you actually built
The first pillar is genuine technical substance. Assessors respond to a clear account of what the company itself has created, as opposed to what it merely consumes from a third party. Proprietary or hard-to-acquire data, models you have trained or meaningfully fine-tuned, novel tooling, non-trivial systems engineering, an original approach to a hard problem, or research-grade work all count as substance. The key contrast is between building and reselling: a company that has engineered something is innovating; a company that pipes a public model through a form is mostly distributing someone else's innovation.
This does not mean you must have trained a foundation model — very few startups do, and that is not the bar. It means you should be able to point to the specific parts of your stack where your team's own work lives. Where is the difficulty? What did you solve that a competent engineer could not assemble in a weekend from public tools? A convincing answer usually references particular components — a data pipeline, an evaluation framework, a retrieval or agent architecture, domain-specific fine-tuning, real integration depth — rather than a general claim that "our AI is better".
Defensibility and the wrapper problem
The second pillar is defensibility, and this is where the "wrapper" question lives. A pure wrapper — a thin interface over a general-purpose model, with no distinctive data, engineering or workflow depth — is the archetypal weak case, because anything a wrapper does can typically be reproduced quickly by others using the same underlying model. That does not make every wrapper hopeless: some evolve into genuinely defensible products through proprietary data accumulation, deep workflow integration, network effects, regulated-domain expertise or systems work that is hard to copy. But the defensibility has to be real and articulable, not a hope that customers will not notice how thin the moat is.
An assessor is implicitly asking a competitive question: if the model provider or a well-funded competitor tried to do this tomorrow, what stops them? Strong answers point to something the founder controls that is not trivially available — unique data, a hard technical achievement, deep domain integration, or an execution advantage backed by real traction. Weak answers describe features that are one prompt-template away from being replicated. Being honest about this internally is also good business, but for the report specifically it is the difference between a file that reads as an innovative venture and one that reads as a repackaging exercise.
Economic interest and scalability
Innovation is only half of the standard; the report also looks for special economic interest. That means the project should credibly matter economically — a real market, a scalable model, the potential to create value, employment or exports, and a business that can grow rather than a science project with no commercial path. AI startups are often well placed here because software scales, but the economic case still has to be made rather than assumed. A genuinely novel technical idea with no plausible market is not obviously of special economic interest, and a strong file connects the innovation to an economic story.
Scalability is part of this. A product whose unit economics improve with scale, that can serve many customers without linear cost, and that has a credible route to meaningful revenue reads as economically interesting. The founder should be able to explain not just what is clever about the technology, but why that cleverness translates into a business worth building in Spain. This is also where investor traction, pilots, revenue or a serious pipeline help — not as proof of innovation, but as evidence that the economic-interest limb is real.
Team and capacity to execute
The fourth pillar is the team. An innovative claim is more credible when the people making it can plausibly deliver it. Technical founders with relevant backgrounds, prior building experience, domain expertise and a coherent plan to execute strengthen the picture; a bold technical vision with no one on the team capable of building it weakens it. Assessors are evaluating a real venture, and real ventures are built by people. The team section is not a formality — it is part of why the innovation and economic interest are believable.
| Signal | Reads as innovative | Reads as AI-washing |
|---|---|---|
| Technical core | Proprietary data, fine-tuning, novel systems work | Thin UI over a public model API |
| Defensibility | Hard to copy — data, depth, integration, traction | Reproducible in a weekend by anyone |
| Economic interest | Real market, scalable model, value/exports | Cool demo, no commercial path |
| Team | Can plausibly build what they describe | Vision with no one able to execute it |
| Evidence | Architecture, roadmap, IP, metrics shown | Buzzwords and adjectives asserted |
Showing it, not asserting it
Across all four pillars, the recurring failure is assertion in place of evidence. "Our platform is highly innovative and uses cutting-edge AI" is a sentence any company can write, which is exactly why it persuades no one assessing a file. The stronger approach shows the innovation: an architecture description that makes the technical contribution visible, a roadmap that reveals depth, a clear statement of what data or models the company controls, IP where it exists, and metrics or traction that evidence the economic case. The same material a good technical founder would show a sophisticated investor is usually the material that supports the report.
This is also where good preparation pays off, because the innovation narrative, the residence file and the eventual tax position all need to tell one consistent story. A project described as deeply innovative for ENISA but structured and priced as if almost nothing of value is created will read inconsistently. Founders should build the evidence base once and use it coherently across the venture. For how the report fits the entrepreneur route mechanically, our ENISA report guide and the entrepreneur versus highly-qualified route comparison set out the surrounding framework.
How this connects to Beckham
The reason any of this matters for relocation is the link to the Beckham Regime. A favourable report supports the innovative-entrepreneur route, and that route is one of the qualifying reasons that can open the special regime under Article 93 — the flat 24% on the general base up to the threshold, and Spanish-source focus for the covered years. For an AI founder moving to Spain, a strong innovation file is therefore not just a startup credential; it can be the gateway to the tax regime, provided the whole picture holds together.
But the report is one piece of a larger file. The corporate structure, any foreign parent, the founder's role and the risk of creating a permanent establishment in Spain all sit alongside it, and a great innovation story does not cure a weak structure. Founders should read this together with the pillar guide on applying for the Beckham Regime in Spain and, if there is a foreign holding company, the note on foreign company owners and permanent establishment. Get the innovation genuinely right, evidence it properly, and make sure the rest of the file tells the same story — that is what turns an "AI startup" into a defensible entrepreneur-route relocation.
Frequently asked questions
Does using AI make my project innovative for ENISA?
No. The report assesses genuine innovation and economic interest, not the technology label. Building a thin interface over a public model is not evidence of innovation on its own. What counts is the technical substance, defensibility and market underneath.
Is an AI wrapper enough to qualify?
Usually not by itself. A pure wrapper with no distinctive data, engineering or workflow depth is the classic weak case because it is easy to reproduce. Some wrappers become defensible through proprietary data or deep integration, but the innovation must be real and documented.
What actually makes an AI project look innovative?
A real technical contribution such as proprietary data, models or systems work; a defensible advantage; a scalable model with genuine economic interest; and a team able to build it — all shown with evidence like architecture, roadmap, IP and traction, not just asserted.
Do I need to have trained my own model?
No. Very few startups do, and that is not the bar. You need to point to where your team's own work lives — data, fine-tuning, tooling, systems, integration depth — and explain what you solved that is not trivially assembled from public tools.
How does the report help with the Beckham Regime?
A favourable report supports the innovative-entrepreneur route, one of the qualifying reasons that can open the Beckham Regime under Article 93. But it is one piece — the corporate structure, your role and permanent-establishment risk all sit alongside it and must be coherent.
General information, not legal, tax or immigration advice. Sources reviewed include the Startup Law 28/2022 and its innovative-entrepreneur framework; the ENISA favourable report used to evidence innovation and special economic interest; Article 93 of the Spanish Personal Income Tax Act (IRPF) and the special regime for displaced workers; and general guidance on permanent establishment and place of effective management.