„It once occurred to a certain king, that if he always knew the right time to begin everything; if he knew who were the right people to listen to, and whom to avoid; and, above all, if he always knew what was the most important thing to do, he would never fail in anything he might undertake.” – Leo Tolstoy, „Three questions” (Three Questions by Leo Tolstoy)

The preliminary question in Case C-649/23 – The “G. Călinescu” Institute of Literary History and Theory[1] was referred by the Romanian High Court of Cassation and Justice to the Court of Justice of the European Union at a moment when the latter had already established most of the reference points relevant to the issue at stake: it had indicated that the requirement for copyright protection in the EU is originality which meant „author’s own intellectual creation” (Infopaq); it had established that originality only required that the object reflect the author’s personality by manifesting the latter’s free and creative choices (Cofemel, Painer); it had clarified that intellectual effort or the application of know-how are not relevant for the purposes of evaluation of originality (Cofemel, Football Dataco, Funke Medien); it had stated that a word can’t be protected as such since the creative spirit of the author can only be manifested through the choice, arrangement and combination of words (Infopaq).

In so doing, the CJEU was perceived to have harmonized the understanding of the condition of originality for the protection of copyright in the EU clearly within the expansive scope of the subjective understanding of originality circumscribing the centrality of the author’s personality as recorded by the author’s free and creative choices.

In this context, the importance of, and justification for, copyright protection seemed to have rested upon an unshakeable belief in the inherent value of promoting (and rewarding) the creation of any material which is novel, which novelty would be guaranteed not by means of an objective measurement of differences against a “state of the art” but rather by the imprinting (by means of said free and creative choices) of a personality which is, by definition, unique. Thus, protection would be justified by, and limited to, the free and creative choices reflecting the author’s unique personality.

Advocate General Spielmann’s Opinion in the case seems, in this context, to raise three important questions left for the Court to answer.

1. The Watch will soon end. Will that be the end of watching it, too?

Termed “Operation Night Watch,” the very ambitious project to study and restore Rembrandt’s 1642 masterpiece painting, “Militia Company of District II under the Command of Captain Frans Banninck Cocq,” known as “The Night Watch”, has begun in 2019 and is still ongoing at the Rijksmuseum in Amsterdam.

Such projects, aimed at restoring old works of art, become, as technology advances, ever more complex in their efforts to learn more about the work prior to restoration. In 1976, for the restoration of Leonardo da Vinci’s “The Last Supper,” examination of the work prior to the actual restoration included sonar and radar tests, and use of infrared cameras and x-rays[2]. For “The Night Watch” it was five years of intensive research including taking the largest and most detailed photo ever taken of a work of art (717 gigapixels in size, with one pixel being smaller than a human red blood cell) made up of 8439 individual photos stitched together by artificial intelligence[3] and use of a micro X-ray fluorescence (Micro-XRF) spectrometer specially developed for the non-invasive on-site analysis of large format objects.[4]

The high level of the technology used to understand the work undertaken by the original author and the effects of previous restorations is impressive. Doubled by the very high level of specialization and expertise of the restorers involved in the process, this comes only to emphasize the unmistakable aim of these projects: to restore the vision of the original author as reflected in the original work. The newest technology, specialists with years of study of the author’s life and work all intensely scrutinizing every millimeter of the work itself for years on end just to make sure that the restoration reflects, as accurately as possible, the original author’s expression.

All this can sum up the context in which the restorer performs her work like this: undisputably, the intention is to expend no effort to make sure that the work reflects to the highest degree possible the expression of the original author, the restorer’s success being measured in how much her own opinions, choices and personality are left aside. All the research and specialization is needed to reduce as much as possible the space wherein there would be insufficient clarity as to the original author’s expression.

It is about that space that AG Spielmann writes “when [the restorer] does not rely solely on her profession and knowledge of the author, language, epoch and of the original work, but also on her own imagination, intuition and sensibility to invent or reinvent, while attempting at the same time to remain faithful to the spirit of the original work, lost or incomprehensible elements, that the author of the critical edition undertakes a creative activity and imprints her own personality upon the work of another, thereby creating a derivative work.”[5]

The reality is that that space is very much there and it is, on most occasions, the very reason for which restorations are needed: if the original work would be today as was created by the original author, no restoration would be necessary and if the original expression was evident to anyone, its mere replication could be undertaken exclusively by mechanical means, no years of studying the work of just that particular author needed.

A great example is da Vinci’s “Last Supper” about which Ross King tells us that the restorers of the 20^th century had to not only rely on their years of studying da Vinci, on Leonardo’s drawings and early copies of the work to reconstruct the faces of the twelve apostles, but also to try to reverse the modifications to the painting of the previous restorers (who, King notes, had made Judas’ figure into a “caricature of evil.”)[6]

In so doing, the restorers clearly had to choose between competing versions (in the various copies) and restore lost or incomprehensible elements, such as those which had been modified by the previous restorers of the work. That the intention was always to remain faithful to the original work is unquestionable. That within that space, not everything the restorers contributed was a direct and necessary result of their application of their professional knowledge is very much possible.

Is thus, “The Last Supper” as last restored by Pinin Brambilla Barcilon between 1978 and 1999, no longer da Vinci’s? Is it, as restored, a derivative work, the copyright in which belongs to Ms. Barcilon? Ms. Barcilon passed away in December 2020, therefore her copyright in the derivative work would last until 1 January 2091, with her estate able to, among others, exclusively authorize the making and distribution of copies of the work, as well as to forbid the public display of the work. The Convent of Santa Maria delle Grazie of Milan, where the work has been publicly displayed since da Vinci finished it in 1498, should take notice.

That in such case the copyright of the restorer would extend to the whole of the work is made clear at par. 70 of AG Spielmann’s Opinion where the “separable nature” of the contributions is dealt with: “We believe that the critical edition of an original work may, in its totality, constitute the object of a precise and objective identification. A distinction need not necessarily be made between the parts corresponding to the original work having eventually suffered textual modifications, on the one side, and the comments, critical notes or explanations accompanying them, on the other, in order to identify accordingly the acts conferring copyright and those that don’t.”[7]

Thus, reading into the Opinion, we could conclude that a restorer who has, while trying her very specialized and professional best to reflect the original author’s intentions, added to the space where such were not absolutely clear what she herself deemed (or guessed) to have been the original author’s intention, also becomes the holder of a brand new copyright in the whole of the restored work, as a derivative work.

So, it appears it is now for the Court to say whether every recently restored work will be within the copyright remit of its restorers to forbid the public display of. Museums around the world should shudder at the thought. But the question has been raised. Now is the time to answer it.

2. Author, Restorer, Visitor. Whose voice to listen to?

The thinking behind the AG’s Opinion does seem solid: establishing that a category of works altogether would not be copyrightable seems to go against the whole idea that protection of copyright should apply extensively to any object meeting the right conditions. This is a direct effect of the reasoning that copyright protection is a means to encourage diversity of creation, irrespective of medium, value or form.

But critical editions or restorations are not really the same as a work, just as a copy isn’t. A copy that, in spite of every effort by the copier not to, includes a mistake that is clearly attributable to the copier would seem to qualify as a derivative work, with the copier achieving, by means of error, what it could not achieve by law: acquiring an exclusive right in the whole of the work, due to his inability of keeping the copy exact.

Without now going into the philosophical exercise of whether the ship of Theseus is still that after every piece of wood has been replaced, our question now would be: if the modern Athenian restorer would be convinced that Theseus had built his ship out of maple and would now replace all the wooden pieces in the ship by identical pieces but made of maple (which, the same restorer would confide, would have never been his wood of choice for building ships), would that choice (made in the spirit of Theseus, not the restorer) now make this no longer “the ship of Theseus” but “the restorer’s ship”?

Of course, the AG’s Opinion does indicate that: “Recognizing a copyright in a derivative literary work does not bring the original work back into the private domain and does not grant the author of this derivative work any exclusive right to the original works. Use of works in the public domain is always possible, these being no longer individually protected by copyright, thus leaving anyone free to copy or even adapt them in a manner different to the derivative work.”[8]

But how would the original ship be viewed now? It is no longer. There is only the derivative work and no one can take the maple out of it. Just like “The Last Supper”: the original work is no more and can’t be used independently of the derivative work. The restorer’s contributions are bound to the original work and can’t be separated anymore nor restored to its version before the last restoration.

Moreover, if copyright in the derivative work is recognized, the only use that can be made of the original work (which was and, according to the AG, remains in the public domain) requires authorization from the copyright holder in the derivative work, from which it can’t be separated.

And thus the visitor of the work is harmed. She loses something she previously had. The author, who had no economic rights in the work and, presumably, can still exercise any moral rights still recognized, has nothing to gain from this, while the restorer gains something she had never sought: a right to something she tried her best not to assert any ownership of.

The other restorers would be harmed as well: had they agreed with some choices made by the initial restorer for the restoration as correctly reflecting the original author’s intentions, but not others, they could not replicate those choices they agreed with in their own restorations while making different suggestions for the ones they didn’t agree with. Thus, new restorations/critical editions would either be at the discretion of the previous restorer/editor or would need to be less actual restorations/critical editions guided by the purpose of most accurately reflecting the intention of the original author and more by the need to avoid infringing the previous restorations/critical editions.

This reflects the basic difference between the works copyright is meant to encourage and restorations/critical editions: copyright is meant to encourage diversity of expression by requiring that personality be bound to it. Restorations and critical editions are meant to bring back works which were already created, expression of personality which has already occurred. There is no more that copyright can do here as these are works which have already been created but are no longer accessible due to time having caused the expression to have become, in some parts, unclear or unrecognizable.

Of course, anyone starting out to creatively reinterpret or adapt an existing work is setting on a path where her personality will guide her expression beyond the boundaries of the pre-existing work and would thus add to the diversity of existing expression, not just find some of the expression already existing and making it visible or accessible again.

So, will the Court heed the author, who has already created and contributed her expression to humanity and has long ago reaped the benefits thereof? The visitor who had (albeit limited) access to this expression and could lose it all without knowing? Or the restorer who has made again available expression which was not hers but was lost and would have remained so but for her efforts?

3. The newly-made original Night Watch. The most important digital elephant in the room.

And what of the process? The Opinion reads: “When an author tries to reconstitute an incomplete literary work to the form she considers closest to that created by the author of the original work, one can’t consider, in principle, that she only contributes with research, a know-how or an intellectual effort lacking all personal creativity. On the contrary, it is possible that the author have the possibility of choosing between more options. Grammar, lexical, literary and stylistic choices that she makes will probably be determined, or at least influenced by, her years of experience, her philological expertise, her knowledge and her understanding of the period when the original work was written and of the historical period the work concerns, her knowledge of the author of the original work, of his style and linguistic expression, as well as her interpretation as to what she perceives to be that author’s intention. When she does not rely solely on her profession and knowledge of the author, language, epoch and of the original work, but also on her own imagination, intuition and sensibility to invent or reinvent, while attempting at the same time to remain faithful to the spirit of the original work, lost or incomprehensible elements, that the author of the critical edition undertakes a creative activity and imprints her own personality upon the work of another, thereby creating a derivative work.”[9]

The Opinion seems to suggest that, where choices are made based on an extensive training on the subject, in order to ‘fill in the gaps’ of an incomplete image according to what is most predictably there depicted, such would not be an effort devoid of personal creativity. This accurately describes the functioning of generative AI models: based on their very extensive training, they try to predict what is most probably the expression that would best fit within the framework established by the person providing it with an incomplete description of what it expects, by means of a prompt.

Although one could argue that, since the model would not use “her own imagination, intuition and sensibility,” the model not having any on account of it not being human, this would seem to shift the argument of no copyright for AI from the qualitative factor of what it is that these models produce to just the fact that only humans can benefit from copyright protection.

This seems even more true when we look at how the AG indicates that originality can be expressed in such a derivative work: “by means of the composition of the critical edition, the structure conferred on the work, the establishing of its form and by arranging the original text by reference to the comments and critical apparatus,” all of which would “impress upon the so-created work a “personal note.””[10]

Prompt any publicly available LLM to create an image based on a descriptive prompt and it will certainly output something which will go beyond the descriptive limits of the prompt in terms of composition, structure or form. Does this mean that the model has impressed its personal note upon the created image? The mere fact that it functions and outputs according to very strict rules would not exclude this, suggests the Opinion at par. 66.

In 1715, less than 75 years after Rembrandt completed “The Night Watch,” the painting was cut in order to fit in between the two doors at the Amsterdam City Hall. Someone trimmed about 60 cm from the left side, 22 cm from the top, 12 cm from the bottom and 7 cm from the right side. These parts were never found.

Using a copy of the whole painting made by Gerrit Lundens and an AI model trained on the works of Rembrandt, “Operation Night Watch” recreated the missing pieces, printed them on canvas and displayed them alongside the original work for the viewers to see, for the first time in over 300 years, what the restorers believed the painting would have originally looked like.[11]

The additions were significant, one report talks of “three figures on a bridge: two militiamen and a young boy. A helmet worn by a militiaman on the extreme right of the painting is completed and a space above the standard emerges, making the motion of the ensign as he raises the company’s colours all the more convincing. It can also be seen more clearly that a boy in the left foreground is grasping a balustrade and running away, ahead of the militia.”[12]

These elements are all reconstructions made mostly by AI ‘translating’ the Lundens copy into what it considers most predictably to have been what Rembrandt painted based on its extensive research of all things Rembrandt. They are, nonetheless, approximations made in that space where the original author’s expression is not entirely clear. They are also a part of the original work as conceived by Rembrandt.

Should the panels be adjoined to what remains of Rembrandt’s painting, would that be a derivative work? That such would have been, at least in part, produced by AI would alter that finding? But surely the added expression, even if carefully and scientifically guided by the aim of replicating Rembrandt’s painting, does involve choices as the ones indicated by AG Spielmann’s Opinion as reflective of a creative effort. Would this allow someone, now, to prevent the public viewing of the newly-completed full restauration of the work as believed to have been created by Rembrandt almost 400 years ago? Using just the panels alone would not make any sense either, what would then be the point of restoring them?

These seem all difficult questions and we do hope that the Court will help us with answers to at least some of them in its decision.

“Remember then: there is only one time that is important – Now! It is the most important time because it is the only time when we have any power. The most necessary man is he with whom you are, for no man knows whether he will ever have dealings with anyone else: and the most important affair is, to do him good, because for that purpose alone was man sent into this life!” – Leo Tolstoy, „Three questions” (Three Questions by Leo Tolstoy)

---

[1] We have represented the “G. Călinescu” Institute of Literary History and Theory in the national proceedings and before the CJEU.
[2] Ross King, Leonardo and the Last Supper, Walker & Company (New York – 2012), pp. 273-274.
[3] https://www.rijksmuseum.nl/en/stories/operation-night-watch/story/ultra-high-resolution-photo
[4] https://www.bruker.com/en/applications/academia-materials-science/art-conservation-archaeology/fine-art-analysis/unveiling-rembrandt-the-night-watch.html
[5] AG Spielmann’s Opinion in case C-649/23, Institutul de Istorie și Teorie Literară „G. Călinescu”, delivered on 26 June 2025, par. 63 (unofficial English translation).
[6] Ross King, op. cit., p. 274.
[7] AG Spielmann’s Opinion in case C-649/23, Institutul de Istorie și Teorie Literară „G. Călinescu”, delivered on 26 June 2025, par. 70 (unofficial English translation).
[8] Idem, par. 75.
[9] Idem, par. 63.
[10] Idem, par. 65.
[11] Caroline Goldstein, “Rembrandt’s Revered ‘Night Watch’ Was Cut Up to Fit Through a Door. With A.I., You Can See It Whole for the First Time in 300 Years”,
https://news.artnet.com/art-world/operation-night-watch-1982686;
[12] Daniel Boffey, “AI helps return Rembrandt’s The Night Watch to original size”, https://www.theguardian.com/artanddesign/2021/jun/23/ai-helps-return-rembrandts-the-night-watch-to-original-size;

---

Paul-George Buta, Managing Partner Mușat & Asociații