There is a version of this argument I want to acknowledge before I make my own. The version goes: all human creativity is derivative, all artists learn by absorbing and internalising the work of those who came before, and training an AI on existing work is simply an automated version of what every art student does in a museum. I understand the argument. I think it is wrong, and I think the people making it know it is wrong.
The art student in the museum looks at paintings. She absorbs technique, composition, colour relationships. She does not reproduce those paintings at commercial scale and sell the reproductions to clients who would otherwise have hired the painters. The distinction between learning from work and commercially exploiting work without compensation is not a technicality. It is the entire foundation of copyright law, which exists precisely because we as societies decided that creative labour deserves the same economic protection as other forms of labour.
The AI companies that built image generators, music composers, and text systems on scraped creative work made a calculated decision. They decided that the legal risk of proceeding without consent was lower than the competitive cost of seeking it. Several of them have said as much in internal documents that have since become public through litigation discovery. They were not confused about what they were doing. They were betting that by the time the legal and regulatory system caught up, the technology would be so embedded in the economy that unwinding it would be politically impossible.
That bet may yet pay off. But the ethical case for it has never been made, because it cannot be made. The beneficiaries of this system are among the most capitalised companies in human history. The people whose work was taken are, in many cases, illustrators and musicians and writers living on margins so thin that losing even a fraction of commercial work to AI competitors has been professionally catastrophic. This is not a conflict between innovation and nostalgia. It is a conflict between capital and labour, dressed up in the language of technological inevitability.
I am not arguing that generative AI should not exist. I am arguing that it should have been built on a foundation of consent and compensation, and that it is not too late to retrofit that foundation through legislation, collective licensing frameworks, and judicial interpretation. The artists did not owe the AI industry their life's work for free. They were simply not asked.