Generative AI Threatens Creative Workers Who Train the Models

Generative AI systems depend entirely on human-created content for training, yet the creative professionals whose work powers these models face an existential economic threat as the technology replaces their labor.

Researchers at the University of Melbourne argue that while copyright enforcement offers some recourse for rights holders, it fails to address the fundamental challenge: generative AI can reproduce creative styles instantly and at scale, undercutting the market for human creators before they can develop their skills or earn a living.

The datafication problem

Generative AI foundation models function by converting vast datasets into tokens—fragments of words, images, or other content—then learning to predict how those tokens should combine in response to prompts. This process represents the latest evolution in datafication: the conversion of human effort into digital information that can be bought and sold.

The researchers acknowledge that remixing existing work has always been part of creative practice. Musicians sample classics to create new genres. Fanfiction writers reimagine established narratives to include perspectives the mainstream industry excludes. But AI-driven remixing differs in two critical ways.

First, AI can recreate nearly any style almost instantaneously, unlike human creators who require time to study and adapt existing work. Second, AI companies can reproduce content at scale without seeking consent, credit, or compensation from the original creators.

Recent legal action illustrates the scope of the issue. AI company Anthropic was ordered to pay $1.5 billion for allegedly using 500,000 books without authorization to train its systems—approximately $3,000 per book. According to the researchers, these settlements address copyright violations but don't solve the underlying labor crisis.

Why it matters

If creative professionals cannot afford to learn their craft and pay their bills while developing expertise, the pipeline of human creativity that feeds AI systems will dry up. The researchers warn that without intervention, AI will continue remixing with no new human input, leaving machines to determine cultural expression.

Beyond copyright protection

The Melbourne researchers propose that companies hosting AI-generated design, music, or voiceover services pay a special levy to firms that employ human creators. This approach would focus on protecting the economic viability of creative work rather than geographic origin or copyright alone.

The threat extends beyond creative industries. Generative AI models are built on the output of workers across sectors, and the datafication process could undermine labor markets wherever human expertise can be tokenized and reproduced.

Australian Prime Minister Anthony Albanese announced a national AI plan last week, though details remain limited. The researchers emphasize that developing AI capabilities and protecting creative workforces are not mutually exclusive goals—both are essential to Australia's economic future.

These findings were first reported by the University of Melbourne's Pursuit publication.