The Tool That Changes the Question
Every major new creative technology — the printing press, photography, recorded music, digital design tools — was initially described as threatening to eliminate the creative professional. Every one of them changed what creative professionals do rather than eliminating them, though the transition created genuine disruption and genuine harm for specific practitioners in specific periods. Generative AI is in the middle of this same transition, which makes it simultaneously genuinely threatening to some working creatives in specific roles and genuinely useful to others who’ve found ways to incorporate it into their practice.
Understanding where AI generative tools are genuinely useful creative instruments, where they’re producing genuine harm for working artists, and where the ethical and aesthetic questions remain legitimately unresolved produces a more accurate picture than either ‘AI is destroying art’ or ‘AI is just another tool like Photoshop.’ Both framings exist because both contain partial truth.
Where Generative AI Genuinely Helps Creative Work
The generative AI applications that working creatives most consistently report as genuinely valuable: rapid ideation and concept exploration (generating a dozen visual concepts from a brief in minutes rather than hours of thumbnail sketching), overcoming blank-page paralysis (a starting point, even an imperfect one, is easier to react to than an empty canvas), and handling the tedious production tasks that aren’t the creative core of the work (background generation, texture creation, format adaptation of assets across different specifications).
Writers using AI for research organisation and rough draft structure, musicians using AI for production elements they don’t have the technical skills to create themselves, filmmakers using AI for storyboard generation and pre-visualisation, and game developers using AI for asset generation all report genuine productivity improvements in the production phases of creative work. These uses don’t replace the creative judgment, direction, and authorship of the human creator — they accelerate specific production tasks.
The Training Data Question That Hasn’t Been Resolved
The most legitimate and most unresolved tension in the generative AI and creativity intersection is training data: the image generation models (Midjourney, Stable Diffusion, DALL-E) were trained on billions of images scraped from the internet without explicit consent from the artists whose work was included, without compensation, and in ways that allow the model to generate images in specific artists’ recognisable styles.
Multiple lawsuits are working through courts in the US and UK challenging the legality of this training approach. The legal outcomes are genuinely uncertain — copyright law wasn’t designed with this use case in mind, and courts are reaching different conclusions in different jurisdictions. The ethical question is somewhat clearer even if the legal one isn’t: using an artist’s entire body of work to train a system that then competes with that artist for commercial work, without compensation or consent, is a genuinely problematic appropriation even if it’s eventually found legal. Artists who’ve found their styles replicated by models trained on their work have a legitimate grievance that ‘it’s just like how humans learn from inspiration’ doesn’t fully address.
The Question of What Counts as Authorship
‘AI-generated’ covers a spectrum from prompting a generative model and accepting its first output unchanged, to using AI tools as one component in a deeply directed creative process where the human makes hundreds of decisions about selection, composition, editing, and integration. The legal and aesthetic treatment of these very different uses as the same category creates confusion in both copyright law (does AI-generated content receive copyright protection, and if not, where on the spectrum does human authorship become sufficient?) and in creative communities (what is the appropriate standard for disclosure of AI use in creative work?).
The disclosure question is the one that creative communities are actively working through: some contexts (editorial illustration, stock photography, awards competitions) have developed disclosure requirements that treat AI-generated and AI-assisted work differently. Others are still in the process of developing norms. The creator who’s transparent about AI tool use in their process is in a better position regardless of how norms develop than one who claims fully human authorship of AI-assisted work — the retroactive discovery of AI use without disclosure is consistently more damaging to creative reputation than proactive transparency.
The Practical Advice for Working Creatives
For creatives navigating AI tools in their practice: the tools with the clearest ethical profile are those trained on licensed or public domain data (Adobe Firefly trains on Adobe Stock with model release, making it safer for commercial use than models with unresolved training data provenance), those used for production assistance rather than for generating the creative core of client work, and those used in ways that the client understands and has consented to if AI use is relevant to the engagement.
The tools with the most contested ethical profile — generating content in a specific living artist’s style, replacing rather than supplementing the creative work that was the value proposition to clients — are also the ones most likely to create reputation risk as norms develop and disclosure expectations increase. Using AI tools thoughtfully, transparently, and in the production phases rather than the creative core phases positions creatives to benefit from the genuine productivity improvements while minimising the ethical and reputational exposure of the more contested uses.