A team of researchers from CNRS, Microsoft Research, and Sorbonne Université has published a systematic analysis of how generative AI is being framed in media coverage of creative work, and the picture they paint is not a neutral one. The paper, led by Baptiste Caramiaux and co-authored by Kate Crawford, Q. Vera Liao, Gonzalo Ramos, and Jenny Williams, argues that the dominant narratives around AI and creativity come overwhelmingly from tech actors and journalists, not from artists. And those narratives carry a specific set of values: that creative ideas should be separated from their material execution, that automation is the primary driver of creative progress, and that removing the need for craft skills is a form of democratization.
The paper is a cultural critique, not a technical one. It does not evaluate whether generative models produce good art. It asks who gets to tell the story of what AI means for art, and what values those stories embed. The answer, the authors write, is that “the discourse promotes creativity freed from its material realisation through human labor.” That is a deliberate choice of framing, not an inevitable consequence of the technology.
The researchers reviewed online media publications returned by popular search engines for queries related to AI and creative work. They found that the voices in those narratives are “primarily not those of artists, but of journalists and tech actors.” When artists do speak on the topic, the paper notes, “a different narrative emerges, one that is contrary to the dominant narrative from tech companies.”
That dominant narrative rests on three interlocking claims. First, that the essence of creativity lies in the idea, not in the execution. Second, that automation is the mechanism that frees the idea from the messy, time-consuming work of making. Third, that removing the skills required for execution makes creativity more accessible to more people.
The paper calls this a “techno-positivist vision.” It is a vision that maps neatly onto the product roadmaps of companies like OpenAI, Adobe, Canva, and Stability AI, all of which market generative tools as ways to “skip the hard parts” of creative production. The message is consistent: prompt in, art out, no training required.
But the paper argues that this framing does more than sell software. It reshapes what counts as creative work. If the idea is everything and the execution is merely a bottleneck to be automated, then the value of craft, of years of practice, of material knowledge, is systematically devalued. The narrative does not just describe a shift. It legitimizes one.
The authors connect this to a broader pattern in AI discourse. They cite the work of Campolo and Crawford on “enchanted determinism,” where AI systems are portrayed as both magical and deterministic, allowing companies to “deflect accountability for the full impacts of their systems.” They also reference the observation by Chubb et al. that dominant AI narratives are “polarized between notions of threat and myopic solutionism.” The creative-work narrative, in this view, is a specific instance of a general tendency: tech companies tell stories that make their products seem inevitable and beneficial, while obscuring the tradeoffs.
One of the paper’s most striking claims is about the relationship between skill removal and democratization. The authors write that “the withdrawal of the skills typically required in the execution of the creative process is then placed in the perspective of a democratization of creativity through the acquisition of rapid skills rather than long-term ones.” This is a subtle but important distinction. The narrative equates accessibility with the elimination of barriers to entry. But what it does not say is that those barriers are also the source of depth, differentiation, and professional identity. A world where anyone can generate an image with a text prompt is not necessarily a world where more people can build a career in the arts.
The paper does not argue that generative AI has no place in creative work. It argues that the current framing is partial and self-serving. “The dominant vision is a partial one, which is primarily not narrated by artists and which, in fact, reinforces artists’ fears and anxieties about this technology,” the authors write. If the goal is to integrate AI into creative practice in a way that enriches rather than diminishes the field, then the narratives need to change.
The paper also situates the current moment in a longer history. It traces AI art back to Harold Cohen’s Aaron software in the 1970s, Vera Molnár’s algorithmic drawings, and Lejaren Hiller’s 1956 Illiac Suite for String Quartet. Generative art is not new. What is new is the scale of the commercial push and the intensity of the media coverage. The earlier experiments were undertaken by artists exploring a new medium. The current wave is driven by companies selling a productivity tool.
For AI builders, the paper raises a practical question. The dominant narrative may be good for short-term adoption, but it may also be creating a backlash. Artists are organizing, lawsuits are piling up, and regulators in the EU and elsewhere are starting to ask whether the data used to train these models was collected fairly. A narrative that frames artists as obsolete is not a sustainable strategy for a technology that depends on cultural legitimacy.
The paper’s value is not in predicting the future of AI art. It is in naming the frame. The story being told about generative AI and creativity is not the only story available. It is the one that serves the interests of the people telling it. The question for the rest of the industry is whether that story is good enough to build on.