Knew or Should Have Known: The Kittel Test and the Burden of Proof

“Should have known” is not a mood of the tax office — it is a legal test the authority must prove from objective circumstances.

“You knew or should have known that your supply chain was connected with VAT fraud” — across the EU, this single sentence decides input VAT deductions, zero-rating and, not rarely, corporate survival. UK readers know it as the Kittel principle behind HMRC assessments; in Germany it operates through Section 25f of the VAT Act. The source is the same CJEU judgment, and so are the safeguards: the knew-or-should-have-known test comes with precise limits and a clear burden of proof. This page sets out the Kittel test, the Mahagében limit on due diligence duties, the evidential standards after Aquila and Global Ink Trade — and why the answer to the allegation is not indignation, but documentation.

The test: what Kittel actually decided

In Kittel (2006), the Court of Justice laid down two sentences that have ordered the field ever since. First, the protection: traders who take every measure that can reasonably be required of them to ensure their transactions are not connected with fraud may rely on the legality of those transactions. Second, the boundary: a trader who knew or should have known that, by its purchase, it was participating in a transaction connected with VAT evasion is treated as a participant in that evasion — and loses the rights concerned.

Two consequences follow. “Should have known” is not a psychological finding but a normative attribution: from objective circumstances and reasonable checking duties, the law infers what a prudent businessperson in that position would have recognised. And the standard is ex ante. What counts is what the trader knew or could have known at the time of the transaction — not what investigators reconstruct years later from databases, mutual-assistance files and seized records. Hindsight bias is the methodological enemy of every defence: what looks obvious today often was not visible then.

Germany codified the test in 2020 in Section 25f of the German VAT Act (UStG): input VAT deduction and the zero-rating of intra-Community supplies are denied where the test is met. What the provision says in detail is covered on the Section 25f page. This page deals with the prior question on which everything turns: when is the allegation proven?

The Mahagében limit: what may be demanded of traders — and what may not

The temptation of enforcement practice is universal: the longer the checklist, the easier the allegation of not having completed it. The Court of Justice blocked precisely that logic in Mahagében and Dávid (2012). The tax administration may not require a trader, as a general matter, to verify that its invoicing supplier possessed the goods, was able to supply them, or had met its own tax obligations. Authorities may not offload their investigative functions onto taxable persons. Enquiry duties arise on cause: where there are indications of irregularity, a prudent trader may be expected to make enquiries about the counterparty — not before.

Germany’s Federal Fiscal Court (Bundesfinanzhof — the supreme tax court) confirmed this line in its 2021 scrap-gold decision: no general duty to examine the supplier’s power of disposal, capacity to supply or tax behaviour; a duty to make enquiries only upon concrete indications; and the burden of establishing constructive knowledge resting with the tax authority. That is the Mahagében limit in German translation — and it deserves the same respect as the Kittel test itself. Sector risk is not suspicion. A low price alone is not constructive knowledge. A trading company is not a tax office.

The burden of proof: what the authority must deliver after Aquila and Global Ink Trade

Here lies the core of this page — and the most frequently overlooked part of the law.

In Aquila Part Prod Com (2022), the Court specified the standard: the tax authority must establish to the requisite legal standard the objective circumstances from which knowledge or constructive knowledge follows. Leading German commentary draws out the consequences: presumptions do not suffice; a reversal of the burden of proof is impermissible because it would impair the effectiveness of EU law; even carousel-style invoicing chains are no more than a serious indication — and the mere fact that participants knew each other proves nothing.

In Global Ink Trade (2024), the Court reinforced the point. The presiding judge of the competent senate of the German Federal Fiscal Court distils the judgment in his annual case-law review: it is for the tax administration to precisely establish the elements of the evasion, to prove the fraudulent conduct, and to demonstrate that the taxable person actively participated or knew or should have known. Three verbs, three duties — establish, prove, demonstrate. Where one is missing, the assessment fails. Rigid administrative checklists are no substitute for that case-by-case examination; national courts must, where necessary, set conflicting administrative guidance aside, because EU law prevails.

That this evidential line is not special to Section 25f is shown by the most recent German scholarship on the parallel rules for incorrect VAT statements: there too, the administration bears the burden of establishing the facts, and estimates must rest on precise, reliable and current data — an analogous reversal of the burden of proof is incompatible with the inquisitorial principles of German tax procedure. The direction of travel is the same everywhere: who alleges, must prove.

And the line keeps moving: in late 2025, in Vaniz, the Court carried the same standard into joint and several liability. A person is to be held liable only where it is established that they knew or should have known that the tax debtor would not pay the tax — here too, the full burden of proof rests with the authority. A few weeks earlier, in Fashion TV RO, the Court had made clear by order that evidence from criminal proceedings against third parties may be used in tax proceedings only where equality of arms and the right to be heard are preserved. The Kittel test thus remains what it is — a rule of evidence. It travels with the allegation, wherever the allegation goes.

The four-step test (who must prove what)

1. VAT evasion somewhere in the chain — to be established and proven: the tax authority.
2. A link between this specific transaction and that evasion — to be proven: the tax authority.
3. Knowledge or constructive knowledge, derived from concrete objective circumstances of this case — to be demonstrated to the requisite legal standard: the tax authority.
4. Reasonable measures — the trader’s exoneration field: whoever can document which checks were run, when, with what result, and how warning signs were assessed, removes the foundation from step three.

Practical reality: the de facto reversal — and the correct response

So much for the law. Now the practice, which deserves a sober look. In German audits and investigations, the authority assembles objective anomalies — price gaps, delivery routes, contact details, sector patterns. Formally, the burden stays with the authority. In practice, the trader must then articulate why it could not have known. German scholarship calls this what it is: a de facto reversal of the burden of proof. And it names the only reliable answer: the counter-evidence generally succeeds only where the trader established suitable control systems beforehand and properly documented the performance of its duties of care.

That diagnosis is no longer a minority view — it is now the position of the commentary literature across camps. Germany’s leading Beck commentary puts it in a single sentence: where sufficient indicia of knowledge exist, the burden of proof is de facto reversed (Heidner, in: Bunjes). And even the commentary closest to the administration openly states that, once the authority has made a substantiated case, it falls to the trader to rebut it — and names the antidote: documented verification (“dokumentierte Vergewisserung”) of the counterparty (Kraeusel, in: Reiß/Kraeusel/Langer). The practical point: whoever wants to counter the reversal needs records, not recollections.

That is not capitulation to enforcement practice — it is its disarmament. A documented checking trail reverses the dynamics: instead of arguing about abstract standards of care, a concrete decision path lies on the table. What was checked? When? With what result? Who decided? Against those questions, the blanket assertion “should have known” is structurally inferior.

A final point is deliberately ambivalent: compliance scholarship is debating a dynamic standard of care — the easier and cheaper data checks become, the harder it is to justify not using them (Teichmann). Anyone deciding today how deep to check should know that current. But the counterweight stands: what remains decisive is the cause-based checking that was reasonable ex ante — not the maximum technically possible effort.

Whose knowledge counts? The two-tier model of knowledge organisation

The allegation is addressed to “the company” — but companies know nothing; people know. The Court made clear in Aquila that a taxable person cannot escape responsibility through delegation: the knowledge of agents and intermediaries can be attributed. All the more important is the counter-question posed by the current German doctoral research on attribution of knowledge: through which organisational channel is the specific knowledge supposed to be attributed? Attribution is a normative operation, not a factual finding about a “corporate brain”.

From this, practice has distilled a two-tier model of due care. Tier one: a baseline control system — the organisational foundation that captures, channels and documents relevant information for every counterparty. Tier two: concrete enquiry duties, triggered only by indications of irregularity. Without a trigger, no boundless investigation of the supply chain; with concrete indicia, form-filling onboarding no longer suffices. And throughout, the timing rule holds: what counts is the knowledge available at the moment of supply — later discoveries must not be reinterpreted as earlier bad faith.

This model is exactly what a VAT-focused tax CMS operationalises: a traffic light system for the tier logic, a Proof of Check for the documentation, cause-based supplier due diligence for tier two. Constructive knowledge is an organisational question — and organisation can be proven.

Fighting VAT fraud is legitimate. But an allegation that has rules of evidence must be measured against them. Those who know the standard can meet it — and prove that they did.

FAQ