On July 16, 2026, the Superior Council for Literary and Artistic Property (CSPLA) issued a report that refuses to create a neighboring right for purely synthetic productions and proposes a three-stage framework for assessing human creative contribution, without requiring proof at all three moments. The report treats traceability as a possible evidentiary indicator, not as a general obligation substituted for a prohibition. This choice has a logic that debates on creative AI have not yet exhausted.

The essentials

  • The CSPLA maintains existing copyright: only an identifiable human contribution opens the right to protection.
  • The contribution can be established at one of the three moments (project conception, assisted generation, selection or final retouching), provided that creative choices appear in the final work.
  • Purely synthetic productions, without documented human intervention, remain outside the scope of copyright.
  • The European Commission imposes transparency and documentation obligations on general-purpose AI model providers, without any requirement that protected works be discoverable.
  • The burden of this proof remains to be defined, particularly for independent creators who lack dedicated resources to manage it.

A market saturated with images without authors

Each month, hundreds of millions of images, texts and music pieces leave servers without a human having held a brush or chosen a word. Content platforms do not distinguish between them. The recommendation algorithm serves indifferently a photograph reworked by a photographer for three hours and an image generated in three seconds. For the human creator, market saturation translates into declining rates and growing invisibility.

Revenues for independent photographers and illustrators have fallen in several countries since 2022, according to sectoral surveys by collective management organizations. UNESCO, in its report Re|Shaping Policies for Creativity published in 2026, reaches the same conclusion on a global scale: the diversity of cultural supply measured in active titles is increasing, but revenue concentration is intensifying.

The classic regulatory response would have been to prohibit or tax. The CSPLA chose differently.

The concrete effects of the CSPLA report

The report of July 16, 2026 starts from a principle that French law has long upheld: only the human mind can create a protectable work. A machine cannot be an author. This foundation is not being called into question.

The report establishes a proof regime adapted to contemporary workflows. A photographer who uses a generative tool to explore compositions, selects one track, retouches the lighting and frames the final result has exercised creative judgment at each stage. Protection can be claimed, but the human creative contribution must be established if the challenge is supported by concrete indicators.

The three moments chosen are deliberately broad. Conception covers upstream choices: the creative brief, the intent, the style sought. Generation covers interaction with the tool: the instructions given, the parameters chosen, the attempts abandoned. Selection or retouching covers final arbitration: what remains, what disappears, what is transformed. At each stage, the author accumulates a trace of their intervention.

This regime does not create new paperwork for creators already working with professional software, which record version histories. It does, however, create a useful asymmetry: entirely automated productions without significant human intervention do not benefit from copyright; the absence of documentary evidence is not, by itself, decisive.

Becker’s thesis tested against data

Adam Becker, in his work More Everything Forever, documents how Silicon Valley’s grand technological promises often mask a logic of capture: generative AI models were trained on corpora made up of human works, sometimes without the consent or compensation of original creators. The value accumulated over generations of artists, writers and musicians has fed systems that now compete with their successors. The promise of democratizing creation conceals a reality of concentration: a few companies control the models, set the conditions of access and capture the value produced.

This reading directly illuminates the CSPLA’s choice. If traceability becomes the condition of the right, it makes visible what was invisible: the human contribution in a hybrid process. It shifts the cursor from production to intent and judgment. It makes it harder, for a large-scale content producer, to simulate an authored work by adding a symbolic retouch to an automatic output.

Becker’s thesis deserves to be confronted with a competing reading. Philippon, in his work on market concentration, distinguishes rent capture by a few actors from the effective diffusion of technological gains to users. The generative tool placed in the hands of an independent illustrator can enable them to explore ten times more options in an hour.

The issue concerns the market structure in which this tool is inscribed. The CSPLA report addresses the question of protection without settling the question of model concentration, which falls under different regulation.

Both readings converge on one point: traceability can constitute an evidentiary indicator, but the report does not subject it to a condition of interoperability with platforms. An author metadata embedded in a file is useless if the recommendation algorithm ignores it. This is precisely what the European Union has imposed transparency obligations regarding content generated or manipulated by AI, not a requirement for discoverability of protected works. The same logic applied to local news, where paying platforms did not suffice to make information more visible.

The cost of proof weighs on independents

The proposed traceability regime is intellectually coherent. Its point of fragility is practical: proving one’s creative contribution requires documenting it, and documentation takes time.

For a creative studio working with integrated professional tools, version histories are automatically generated. For a freelance illustrator using multiple tools in parallel, or a musician moving from one application to another, documentation is an additional burden. Collective management organizations can play a role here: some have begun developing simplified deposit tools, modeled on existing work registries. But deployment remains partial.

Copyright has always had this structural flaw: it protects better those who have the resources to invoke it. A creator who knows how to write a contract and keep archives is better equipped than a beginner publishing directly on a platform. Three-moment traceability reproduces this bias if it is not accompanied by accessible tools. This asymmetry between well-equipped creators and isolated creators echoes the one observed in collective bargaining over AI at work: large structures adapt quickly, independents wait.

The CSPLA report acknowledges this implicitly by calling for support measures. Concrete implementation will depend on choices by the Ministry of Culture and professional organizations.

The trade-offs the attention market will impose

The 2026-2035 horizon raises a question broader than copyright alone. The volume of automatically generated content is growing faster than the capacity of algorithms to qualify it. Platforms have a commercial interest in distributing abundant and inexpensive content. They have an opposite, less immediate interest in maintaining user trust in the quality of what they consume.

This second interest could, over the medium term, create a spontaneous demand for traceability. A music streaming service that signals entirely human compositions to its most loyal subscribers is already testing a differentiation signal. Several photo platforms have introduced filters allowing users to exclude generated images from search results. This movement is still marginal, but it follows an identifiable market logic: if users value authenticity, platforms have an interest in making it readable.

The optimistic scenario is one of convergence between French regulation, European discoverability requirements and the evolution of user preferences. In this scenario, traceability becomes a de facto norm, supported by the platforms’ own tools, and the cost of proof for creators eases as software incorporates it natively.

The less favorable scenario is one of fragmentation: large American platforms apply their own standards, incompatible with European metadata. Creators exporting their works outside the Union end up in a legal vacuum. Traceability protects within the internal market but disappears at the border.

This scenario already corresponds to the current situation for press neighboring rights, whose application to American actors remains incomplete.

Between these two trajectories, the signal to watch is whether or not large platforms adopt open standards for author metadata. The Coalition for Content Provenance and Authenticity, which brings together Adobe, Microsoft and several media companies, has developed such a standard. Its adoption by mass distribution platforms remains, in summer 2026, very partial.

France’s position in a debate playing out elsewhere

France is not alone on this issue, but it is moving faster than most of its partners. The United Kingdom has opened a consultation on a data extraction right for AI training that goes in a different direction, toward mandatory licensing rather than traceability. The United States does not yet have a federal framework. The European Union has set markers in the AI Act, notably the obligation for general-purpose AI model providers to publish a sufficiently detailed summary of the contents used to train them, but without going as far as the proof regime for creative contribution.

This French advance can transform into an exportable standard or remain an isolated exception. It will become a standard if the European Commission takes it up in sectoral texts on digital copyright expected by 2027. It will remain an exception if each member state builds its own regime, making creator protection as fragmented as the digital market itself.

Creator organizations have understood the stakes: SACD and SACEM are among the consulted stakeholders; the report does not document common support from SACD, SCAM and SACEM for its orientation, while calling for support measures to lighten the burden of proof for individual authors. They have also asked that remuneration for the use of works in training corpora be settled in a separate text. This second workstream remains open.

Actions available to creators right now

The CSPLA report does not yet apply: it proposes a method for applying existing law and examines certain possible legislative developments. But it gives clear guidance to creators working today with generative tools.

Documenting the creative process is not a new constraint: it is a practice that illustrators, photographers and architects have always had, in the form of sketches, mockups and intermediate versions. The difficulty is transposing this practice to digital tools that do not always preserve history by default.

Several studios have begun integrating process journals into their workflows: a dated note at each decisive step, automatically exported from the software. This type of documentation serves both legal protection and dialogue with clients, who better understand what they are paying for when they see the path taken.

Traceability, in this context, ceases to be a regulatory constraint to become a commercial argument. The same movement is observed in other sectors where AI enters skilled work: value shifts toward what the machine cannot document alone—judgment, intent, choice.

The real test of the CSPLA report will come when the first court conflict over a hybrid work is settled on the basis of its recommendations. That moment will tell whether the three moments of proof constitute a workable framework or an excessive burden. Until then, creators who document their process today are better prepared than those who wait.

Sources

  1. Ministry of Culture, Superior Council for Literary and Artistic Property (CSPLA), Report on Mission on the Status of Artificial Intelligence Productions, July 16, 2026, culture.gouv.fr
  2. Adam Becker, More Everything Forever: AI Overlords, Space Empires, and Silicon Valley’s Crusade to Control the Fate of Humanity, scientificamerican.com
  3. UNESCO, Re|Shaping Policies for Creativity, 2026, UNESCO (URL not verified)
  4. Thomas Philippon, The Great Reversal, work on digital market concentration, Princeton University Press
  5. Coalition for Content Provenance and Authenticity (C2PA), open author metadata standards, c2pa.org