Naming an artist in a prompt
The gesture has become trivial. Type three words, add a name, click. "In the style of" followed by the surname of a working artist brings up an image that looks, sometimes uncomfortably, like what that artist makes. The ease of it hides what it is. A living artist, someone who still works, still sells, still builds the value of their name, has part of what makes them singular put to use without consent, every time a request contains that name. Copyright law as it applies to AI and record labels doesn't address the question head-on: courts are reluctant to accept that an art style can be copyrighted, and French law takes the same line. The act remains, at the very least, inelegant. Sometimes it's worse.
What's at stake isn't only the reproduction of a line, a palette, a texture. It's the capture of a colleague's or a predecessor's professional identity, reduced to a string of characters anyone can type, anywhere. The line between cultural reference and aesthetic predation gets drawn here, and it depends less on the result than on the stance of the person asking.
What copyright recognizes, and what it ignores
Copyright protects works, not styles. The rule is stable in France and in most civil-law countries, and it's a healthy one. It lets one artist draw on another without paying royalties, and it stops an aesthetic movement from being monopolized by whoever formalized it first. Without it, modern painting would be legally impossible. Picasso took from Cézanne, who took from someone else, and so on, all the way back.
That legal architecture was designed for a world where imitating a style took work. Pastiche was slow, deliberate, risky, and because it cost effort, it committed its author. By the act of imitating, the imitator became an interpreter. Generative AI abolishes the effort. It delivers the resemblance without the commitment, the line without the hand, the signature without the accountability. Collective rights bodies such as SACD, the French society of dramatic authors and composers, have started pressing lawmakers in public about the gap between an inherited framework and the way generative models are actually used.
The artist's name as an ingredient
When a prompt uses the name of a living artist, the name works like an ingredient. Inside the model it points to a statistical function, a set of parameters tied to a body of images that artist made, absorbed during training. The artist is no longer a person. They've become a tag, a shortcut, a label in a catalog. That reduction is the exact opposite of what an artistic signature means: a gesture that can't be reduced to any other known gesture.
For living artists whose names sit in the training data, this isn't vanity. It touches their income, if clients can get for a few cents what they used to pay several thousand euros for. And it touches their professional identity, if the singularity they built becomes a side effect of anyone's query. The class action brought in the United States by illustrators including Sarah Andersen and Karla Ortiz against Stability AI put the question in public, though no final judgment has settled the merits.
Three lines not to cross
With no clear legal rule, three ethical thresholds seem workable for a professional who uses generative tools and doesn't want to slide into parasitic capture. First: don't name a living artist in a prompt meant for commercial work. Historical distance, a generation or two, changes the nature of the act; invoking Moebius or Saul Bass is not the same as invoking a colleague still active on the market. Second: don't publish an image recognizable as part of an identified artist's vocabulary. Pastiche, if you practice it, stays private, or says what it is. Third: prefer prompts that describe an intention over prompts that cite a source. Write what you're looking for, not the name of the person who already found it.
Style as a silent deposit
Still, having no rights over a style doesn't mean it has no value. A style is a deposit, the often invisible accumulation of decades of decisions, mistakes, second attempts. It belongs to no one in the legal sense, and to its author in the strongest moral sense. Treating it as a resource free for the taking because no penalty will follow comes from the same reflex that says what can't be seen doesn't count. The reflex is typical of periods when technology runs ahead of moral refinement.
Some professional platforms, in music as in imagery, have begun banning in their terms of service any prompt that explicitly names a living artist, and the World Intellectual Property Organization coordinates working groups on how existing law relates to identifiable styles. These shifts are slow and piecemeal. They do signal that the problem won't be ignored forever. In the meantime the responsibility sits with the creator, with their ability to remember, at the moment of prompting, that there's still someone at the other end of the name they're about to type.
Using generative models with judgment means not feeding on the work of your contemporaries. More on creative AI.