There is a loss that annihilates the value of a work without touching it. A panel sold as autograph, downgraded by a catalogue raisonné committee to workshop production, loses most of its price overnight. The material has not moved by a micron: the panel, the ground, the paint layer are exactly what they were the day before. What changed is an opinion, and that opinion is worth several million. It is the most poorly understood boundary in the line, and it becomes clearer once one accepts from the outset that what is insured here is not an object but the market's agreement about what that object is.
No damage cover responds, and it is worth seeing why the wording deceives nobody once read. Policies in this line readily present themselves as covering all risks of physical loss or damage. The word that decides is not 'all risks', it is 'physical'. A de-attribution produces neither physical loss nor physical damage: it produces pure financial loss, and the policy covers none. The event that destroys the most value in the art market is therefore foreign to that market's principal cover, and the surprise this causes comes entirely from the name given to the product.
Then comes an asymmetry that has to be stated plainly because it is counterintuitive and decides files. Agreed value is a convention, and a convention does not evaporate because the world changed its mind. A work agreed at three million, downgraded to four hundred thousand on the market, still carries an agreed value of three million so long as the contract has not been revised. If it burns, the insurer owes three million for an object worth four hundred thousand. The destruction of a de-attributed work is therefore fully insured while the de-attribution itself is not insured at all, and that gap is not a drafting defect, it is the mechanical consequence of two rules that do not speak to each other.
This is exactly where disclosure enters, and it is the second place where two rules meet. A de-attribution is a material change in what the insurer bound itself to. Must the insured disclose it? If they do, the agreed value will be revised downward and they lose what the convention guaranteed them. If they do not, they leave in place a gap the insurer will discover at the first loss, with the consequences contract law attaches to an inaccurate disclosure. The answer depends on the nature of the contract as much as on its wording: an annual contract puts the question at each renewal, a continuous one puts it when the fact arises, and the two do not leave the same room.
The buyer's recourse is therefore not an insurance recourse but a contractual one, and it runs into time. The dealer who sold the work generally owes a warranty of authenticity, whose scope and duration appear on the invoice or in the conditions of sale. That is where the remedy sits, not on the side of the damage policy. But those warranties are bounded, often to a few years, and the difficulty is structural: a de-attribution rarely arrives quickly. It arrives when a committee meets, when an analytical technique advances, when a researcher reopens an archive, which is to say decades later, long after the warranty has expired.
One calendar feature makes the thing crueller still. The market re-examines a work when it moves. Committees, technical analysis and provenance research converge at the moment of a sale, because that is the moment someone has a reason to pay to know. The collapse in value is therefore triggered by the very act of realizing that value: one learns the work is not worth what one thought at exactly the moment one tries to exchange it for money. A collection that never moves never suffers a de-attribution, which is obviously not the same thing as being authentic.
What remains to be done is modest and does not look like a cover, but it is all there is. On the underwriting side, document the basis of the attribution at the moment the value is agreed, and provide expressly for revision, which stops a gap of several million from living inside a contract without anyone having intended it. On the collector's side, keep the seller's warranty, its date and its scope, and know that it, and not the policy, will carry the subject. Neither discipline repairs a de-attribution. They merely avoid discovering it at the same time as its contractual consequences, which is already most of what can be obtained.
A panel is bought in 2019 from a dealer for 3.2 million euros, with an expert certificate attributing the work to a master. The collector insures it that same year for an agreed value of 3 million, never revised since. In March 2025, the catalogue raisonné committee downgrades the work to workshop production; dealers consulted now value it at around 400,000 euros. The collector does not inform the insurer. In June 2025, an accidental fire destroys the store where the work was kept. The collector claims 3 million. The insurer objects that the work was only worth 400,000.
The insurer's objection is the obvious answer and it is wrong on the ground it has chosen. Agreed value is a convention, it governs total loss, and it was not revised: on the question of value alone, the insurer owes 3 million, exactly as it would have owed them had the artist's standing merely fallen. The competent professional errs here by believing that real value reopens the debate, when the convention exists precisely to close it. The real ground lies elsewhere, and it is disclosure: the March de-attribution is a material change in the risk, and the insurer must examine whether the contract required it to be disclosed and what the silence entails. That ground is uncertain and depends on the contract, but it is the only one that can produce anything. That leaves the file seen plainly, without detour: in March the collector suffered a 2.6 million loss no policy covered, and the June fire returns it in full. The coincidence proves nothing and accuses nobody, but it is a fact any insurer will examine, and it is the reason revising an agreed value is not an administrative formality: it also protects the insured, by removing the gap that makes their own claim look questionable. Finally, the recourse that should have existed is extinguished: the dealer's 2019 warranty of authenticity is out of time by 2025, which is the normal calendar of these matters rather than particular misfortune.
- 01A de-attribution produces no physical damage, so no all-risks damage policy responds: the deciding word is 'physical', not 'all risks'.
- 02Agreed value survives a de-attribution until revised: destroying a de-attributed work is indemnified, the de-attribution itself is not.
- 03That gap is closed by disclosure and revision, not by a cover: it is the only ground on which an insurer can act.
- 04The real recourse is the seller's warranty of authenticity, and it is almost always time-barred when the de-attribution arrives.
- 05The market re-examines a work when it moves: value collapses at the precise moment one tries to realize it.