AI Art and the Authorship Question: Who Made This?

🖼️ The Argument We Are Actually Having

"Is AI art really art" is the question everyone asks, and it is the least useful one available. It has no test attached. Nobody can lose it, because "art" has resisted definition for a century and was already stretched to cover a urinal, a blank canvas and a banana taped to a wall.

The question with actual stakes is narrower: when an image is generated, who is the author?

That one has consequences. It determines who gets paid, who gets credit, who can claim copyright, and who is responsible when something goes wrong. It is also, unusually, a question the art world had already spent a century preparing for without realising it.

🎨 The Preparation Nobody Noticed

Art has been dismantling the idea of the artist's hand for a long time.

Photography did it first: an image made by a machine capturing light, which took decades to be accepted as authored rather than merely recorded. Then readymades, where the artist selected rather than made. Then screenprinting and studio production, where assistants executed the work. Then conceptual art, where the instruction was the piece and the execution was incidental.

By the time image generators arrived, "the artist did not personally make this object" had already stopped being disqualifying. What is genuinely new is not the absence of the hand. It is the absence of a specific, identifiable intention behind each mark.

🧠 Where the Authorship Actually Sits

Rather than one answer, there are four candidates, and the honest position is that it is distributed unevenly between them.

The third row is where the real dispute lives, and it is not philosophical. It is a question about consent and payment that is currently being fought in courts rather than galleries.

⚖️ The Part That Is Genuinely New

Two things are actually unprecedented, and they get lost in the noise.

The style of a living artist can be requested by name. Not imitated by a person who studied it for years, which has always happened and has always been permitted, but produced on demand by a system that absorbed their catalogue. The distinction between influence and extraction has never had to be drawn this precisely before, and we do not have good language for it yet.

The volume changes the economics. A single competent imitator was never a threat to an illustrator's livelihood. Unlimited instant output at near-zero cost is a different proposition. The objection is frequently dismissed as fear of technology when it is a fairly ordinary observation about supply.

🔎 What Has Not Changed

Some things people claim are new are not.

Selection has always been authorship. Photographers take hundreds of frames and choose one. Editors cut. The person who chose which output to keep and which to discard has done something real, even if the thing they chose was generated.

Technical difficulty was never the measure. The idea that art requires visible effort has been repeatedly abandoned and repeatedly revived by whoever feels threatened by the newest tool. It was said about photography, about synthesisers, about digital painting.

Most output is boring, and that is normal. The overwhelming majority of anything is unremarkable. The existence of a great deal of mediocre generated imagery says nothing about the ceiling, in the same way that most photographs being forgettable says nothing about photography.

👁️ The Anonymity Parallel

There is a useful comparison here with Banksy's anonymity.

Banksy's work demonstrates that art can function perfectly well with no known author. You cannot ask him what he meant, and the pieces are not diminished by it. If anything, removing the biography concentrates attention on the work.

That looks like it should settle the AI question in favour of not caring who made it. But the two cases differ in one decisive respect: with Banksy there is an author who is withheld. With a generated image there may be no authorial intention to withhold.

The Banksy piece was made by someone who decided what it should mean and where it should go. The identity is hidden, the intention is not. A generated image can arrive with a plausible surface and nothing underneath, which is why it can feel hollow on second look even when it is technically impressive.

That is the actual difference, and it is not about effort or tools. It is about whether anything was being said.

🧭 How to Think About It Without Picking a Team

If you use these tools, or look at what they produce, a few positions that hold up better than the slogans:

1. Judge the image, disclose the method. Both, not one. Work stands on its own, and audiences are entitled to know what they are looking at.
2. Treat named-artist prompting as different in kind. Making something in a style is old. Requesting a living person's style by name is not the same act.
3. Notice whether there is an idea under it. The most reliable weakness of generated work is that it has surface without argument. That is a real critical test, and it applies to human work too.
4. Do not confuse the tool question with the labour question. Whether it counts as art and whether it is displacing people who need to eat are separate. Both matter, and conflating them lets each be used to dismiss the other.

🔮 Where This Lands

The likeliest outcome is unglamorous: generated images become another medium with their own conventions, disclosure norms and a body of law about training data, and the panic reads as strangely as the panic about photography does now.

What will not resolve is the authorship question, because it was never fully resolved for anything else either. Every medium has an unclear boundary between the person, the tool and the tradition they drew on. Generation made that boundary visible rather than creating it.

The uncomfortable possibility worth sitting with: if a great deal of imagery can be produced with no author and most people cannot tell, that says something about how much of it never required an author. That is not a comfortable thought about machines. It is a fairly pointed one about us.

✅ The Short Version

• "Is it art" has no test attached. "Who is the author" has consequences, and that is the real argument.
• Art spent a century separating the artist from the hand. What is new is not the absent hand but the absent intention.
• Authorship is distributed across the prompter, the developers and the training-data artists. The third is where the genuine dispute is.
• Genuinely new: requesting a living artist's style by name, and the volume changing the economics.
• Not new: selection as authorship, and difficulty never having been the measure.
• Anonymity withholds an author. Generation may have none. That is the difference that matters.