The Philosophy of AI Art: Who Really Creates?
When a machine produces an image that stops you mid-scroll, who deserves the credit? The person who typed the prompt, the engineers who trained the model, or the algorithm itself? This question sits at the centre of the philosophy of AI art, and it has moved from academic journals into everyday creative conversation. For digital artists who now work with generative tools, the answer shapes how they present their work, how they price it, and how they understand their own creative identity.
What Do We Mean by AI Art?
The philosophy of AI art starts with a definitional problem. In the 2023 thesis Art-ificial: The Philosophy of AI Art, AC Helliwell begins by establishing what we mean by “AI art”, examining both examples of AI works and the technological underpinnings of these systems. The term covers a great deal: fully generated images, hybrid human-machine works, and assisted workflows in which AI performs a narrow task inside a larger artistic process.
Most public debate centres on generative models that produce images from text prompts. These systems are trained on vast collections of human-created imagery, learning statistical relationships between words and visual features. When a user types a description, the system draws on those patterns to produce something new. Understanding that process is essential before we can ask who really creates.
Why the Question of Authorship Matters
Philosophers have debated whether artificial intelligence can create art for some time, but the discussion took an interesting turn in 2022, when AI-generated works began winning public competitions. Dr Milena Ivanova, writing for the Leverhulme Centre for the Future of Intelligence in 2025, describes that moment as a shift in the conversation. The debate was no longer hypothetical. Artists suddenly had to reckon with machines that could produce competent images in seconds.
The authorship question carries real weight. If the AI is the creator, the human operator is merely a director. If the human is the creator, the AI is a tool like a paintbrush, albeit an unusually complex one. Between those two positions lies a more complicated middle ground.
The Freedom Argument
One influential approach draws on Kantian aesthetics. In An Aesthetic-Philosophical Analysis of Whether AI Can Create Art, David Winter explains that, according to the Kantian aesthetic approach, art requires freedom. The ability of AI to create art therefore depends on whether we can attribute freedom to AI tools. If a system acts purely according to programmed rules and training data, without choice in any meaningful sense, then its output cannot be free in the way art demands.
This position does not disqualify AI-assisted work. If the human artist exercises freedom in selecting prompts, choosing among outputs, editing results, and deciding when a piece is finished, then freedom is present in the creative loop. The machine contributes fluency; the human contributes judgement.
The Heideggerian Critique
A more sceptical position comes from Karl Kraatz and Shi-Ting Xie in their paper Why AI Art Is Not Art: A Heideggerian Critique. They offer a philosophical explanation for the negative bias many people feel toward AI art. That bias, they argue, rests on our shared understanding of the ontological differences between objects. A painting by a human hand belongs to a different category of being than an image generated by statistical prediction from a database.
The critique does not claim that AI images are worthless. Rather, it claims that they occupy a different ontological status. They may look like art, but they do not emerge from the same relationship between being, world, and expression that human art does. For viewers who hold this view, the discomfort they feel is not prejudice but recognition of a genuine difference.
Mass AI Art and Minimal Human Involvement
Ted Nannicelli introduces the concept of “mass AI-art” in a 2025 paper for The Journal of Aesthetics and Art Criticism. Mass AI-art is characterised by minimal human involvement and is produced in the absence of knowledge of how the AI works. When someone clicks a button and receives an image without understanding the system, refining the process, or intervening in the output, the result is mass AI-art.
Nannicelli’s perspective is moderately sceptical rather than dismissive. The problem is not the technology itself but the way it is used. Minimal human involvement and technical ignorance produce objects that are difficult to call art in any strong sense. This gives critics a benchmark: the more human judgement enters the process, the stronger the claim to authorship.
The Moral Dimension of Training Data
The philosophy of AI art is not only about definitions; it also raises moral issues. In a November 2023 roundtable published by Aesthetics for Birds, eight scholars discussed the fact that many AI systems are trained on human-created samples. This training raises questions about consent, attribution, and fair recompense for the artists whose work feeds the model.
These moral questions connect directly to authorship. If an AI model draws its style from identifiable living artists, then the question of who really creates becomes more complicated. The output is a remix of human labour, and human sources deserve recognition. Artists working with AI therefore operate in a field where the ethical ground is still being settled.
The Artist in the Loop
Hutan Ashrafian of Imperial College London has suggested that the philosophy of art can offer something valuable to our understanding of AI and consciousness. Centuries of thinking about intention, expression, and interpretation may help us understand what machines are doing when they generate images. Those analytical tools remain useful.
For the practising artist, the practical answer is that the human remains central. The artist sets the direction, makes aesthetic decisions, corrects errors, and takes responsibility for the finished piece. A common view in public discussions holds that unless AI becomes fully self-aware and capable of self-expression, its output is not art but a facsimile. That instinct reflects the belief that authorship requires consciousness.
Where the Debate Leaves Us
The philosophy of AI art does not offer a single tidy answer, and that is part of its value. If art requires freedom in the Kantian sense, then the machine cannot create alone; the human operator must supply that freedom. If the ontological critique is correct, then the object produced by an AI has a different status from a human-made artwork, regardless of how it looks. If mass AI-art is defined by minimal human involvement, then the remedy is for artists to involve themselves deeply in the process.
What emerges is a position most working artists can accept: AI is a medium, not a mind. The algorithm generates possibilities, but the artist chooses. The prompt is the beginning of a collaboration with a statistical system, and the meaning of the finished work still depends on human intention and human reception. For buyers and viewers, the question of who really creates may matter less than whether the work moves them.
The debate will continue as the technology evolves. Philosophers will keep refining definitions, and artists will keep testing boundaries. What seems clear is that AI has not erased the artist. It has changed the tools artists use, and it has forced a useful question: what part of making art is irreducibly human?
Frequently Asked Questions
Is AI-generated art really art?
The answer depends on the definition of art you adopt. If art requires freedom and intention, then a purely generated image has a weak claim. If art is judged by the viewer’s experience, the image may qualify. Most philosophers who study the question take a middle position, treating the human operator as the artist and the AI as a sophisticated tool rather than an independent creator.
Can an AI system be the author of an artwork?
Under current philosophical arguments, authorship requires freedom, intention, and self-expression, none of which can be confidently attributed to present AI systems. The Heideggerian critique goes further and suggests AI images belong to a different ontological category from human art. Until AI demonstrates something like self-awareness, most philosophers hesitate to grant it authorship.
Does using AI text-to-image tools make me less of an artist?
Not necessarily. The strength of an artist’s claim depends on how much human judgement enters the process. Selecting prompts, curating outputs, editing results, and combining AI work with other techniques all involve creative decisions. By contrast, mass AI-art, produced with minimal human involvement and little understanding of the system, has a much weaker claim to authorship.
What are the main ethical concerns with AI art?
The central ethical concern is training data. Many AI systems learn from human-created samples, which raises questions about consent, attribution, and fair recompense. The moral debate overlaps with authorship because if the AI draws from identifiable artists, those humans have a stake in the output. Artists using AI should stay informed as these norms develop.
https://www.ArtPal.com/gedespi
FINE ART AMERICA
————————
https://x-gedespi.pixels.com
