The other day I came across a conversation on Reddit about the ethical use of generative AI. People were worried: will we ever have a model built ethically? One that has not taken anything from anyone?
Then someone said the obvious thing the whole discussion was avoiding: no, that is practically impossible. Today’s AI runs on debt. It owes. A good model depends on how much of the artists’ work it captures, and if it had to pay for it, it would collapse on the spot.
And there came the detail almost everyone forgets: the solution to ethical AI is not in where it will learn. It is in how it will be used after learning.
The tool already exists. LoRA exists. ControlNet exists. IP Adapter exists. Even if the model was built on appropriation without consent, it is possible, today, to extract something authorial from it. Something that is yours.
There are two ways to make generative AI authorial and ethical.
One is unfeasible: waiting for a fully ethical model, built from scratch, owing nothing to anyone. That is not going to happen. And if it does, it will not be good.
The other makes sense: legislation that forces models to be built with open doors. Receptive to ControlNet. Receptive to LoRAs. Receptive to IP Adapter. And professional use going through those measures, with clear parameters.
Then the use of AI becomes ethical, not at the origin, but in practice.
And the most beautiful part: this alternative is viable today.
Because ethics, in the end, is not what the machine learned. It is what you do with it.
#AIethics #ArtificialIntelligence #DigitalArt #Creativity #TheAIArtist
