Essential to the profitability and widespread adoption of modern technologies is the need to conceal their costs — the human labour, the unjust or exploitative working conditions, the environmental degradation and drain on natural resources, the reduction of precious goods to mere ‘raw material’. By means of such concealment, those who enjoy the conveniences these technologies afford are lulled into a state of (what Hannah Arendt called) “unthinkingness”. The benefits simply appear, and consumers are left incurious about how it could have happened so quickly, so cheaply, with so little friction.

The other thing that is crucial for the success of such technologies is the way that their speed, financial backing and sheer scale project a sense of irresistibility — what Shoshana Zuboff has termed “the cult of inevitabilism”. Put otherwise, the belief that resistance, however principled and noble, is ultimately futile. So get on board or get out of the way.

Artificial intelligence represents, in many ways, the apogee of both aspects. The assumption that AI development hovers somewhere between an “arms race” with existential stakes and the key to unlocking untold productivity and economic growth has had the effect of reducing calls for prudential care and democratic consultation to little more than inconveniences to be gotten around rather than objections to be engaged with meaningfully. At the same time, the twin mantras of “more compute” (more chips, more data centres, more electricity) and “more data sets” have been used to legitimate both the expenditure of unimaginable sums of money and the indiscriminate (and ultimately illegal) use of intellectual property for the training of AI models. You could call this theft generative AI’s original sin.

In both cases, the guiding philosophy seems to be that the purportedly beneficial ends will have justified the unprincipled means. Which is to say, the emergence of artificial general intelligence will cicatrise the moral wound inflicted by these companies’ originary theft. After all, what’s a little copyright infringement between inhabitants of an AI-ushered utopia?

And yet, beginning in 2024, despite the size, the rapaciousness and the heedlessness of these AI companies, a number of courageous pockets of profoundly human opposition have emerged. To date, perhaps the most consequential are the resistance on the part of local councils in the US and UK to the construction of data centres, and the lawsuits filed on behalf of writers, artists and some media companies against the illegal use of their copyrighted work to train generative AI models.

Met with such opposition, unsurprisingly, AI companies are looking for new terrain on which to construct their data centres, and new sources of data to be strip-mined for their models. Concerns are that the Australian federal government is attempting to position itself as an attractive destination for these companies, and Australia as the beneficiary of the investments that come with them. But at what price to Australia’s creative industries?

Last year, the Productivity Commission proposed a “text and data mining exception” to the Copyright Act, paving the way for AI models to be trained on the copyrighted work of Australian authors and artists. Concerned that this exemption may be revived as part of a proposed deal with AI companies to “attract more than $50 billion worth of datacentre investment”, authors and artists converged on Canberra last week to oppose any such dilution of legal protections to their intellectual property.

The questions raised by this opposition are difficult and extremely consequential. Is “fair compensation” for the use of these artists’ works sufficient to heal the moral wound inflicted by the original theft of their work? Is the establishment of a fund to support future artists enough to encourage and sustain such creative work, when the cultural effect of AI is to devalue both human creativity and the time and labour required to truly create? Aren’t both current and future authors and artists jeopardised by the indiscriminacy with which training models treat “data”, effectively equating the collected works of Dostoevsky with 100,000 hours of transcribed YouTube videos, and reducing them all to raw materials?

• “Search LibGen, the Pirated-Books Database That Meta Used to Train AI”, by Alex Reisner, The Atlantic (20 March 2025).
• “I don’t want a US tech bro as a patron — which is why artists must defend our copyright in the age of AI”, by Anna Funder, The Guardian (6 July 2026).

Guest: Anna Funder is Professor of Practice in Creative Writing at the University of Sydney, and the award-winning author of Stasiland, All That I Am and, most recently, Wifedom: Mrs Orwell’s Invisible Life.