Australia Draws a Line on AI Training. What It Means for Authors

AI & THE AUTHOR

The weekday briefing on artificial intelligence, writing and publishing from the perspective of a working author.

Quick Shortcuts

The story • Why writers should care • The line between AI-assisted and AI-authored • What a fair system could look like • Discussion question

Australia Just Drew a Line on AI Training. Authors Should Pay Attention

For writers, one of the most important AI stories this week is not another new model or another clever writing tool. It is a government saying that copyright still matters even when the technology asking for access is powerful, wealthy and promising investment.

On 22 September, Reuters reported that OpenAI and Anthropic had urged the Australian government to reconsider its refusal to create a broad copyright exemption for training AI models on Australian creative work. Australia has so far ruled out such an exemption, while the companies are arguing for a more limited framework that would allow training under defined conditions.

That sounds like a technical policy dispute. For authors, it is much more fundamental. It asks whether our books, articles, illustrations and other creative work should be treated as raw material that technology companies may use first and negotiate over later, or as intellectual property whose use begins with permission.

Why this matters to writers

The AI copyright argument has spent much of the last few years trapped between two extremes. One side says training is transformative and essential to innovation. The other says creative work has been copied at enormous scale without meaningful consent. Court cases are now testing parts of that argument, but policy decisions may shape the market just as profoundly.

Australia is interesting because it exposes the choice clearly. A country can decide that attracting AI investment requires weakening existing copyright protections. Or it can say that innovation must develop around those protections through licensing, consent and negotiated access.

For authors, that distinction is enormous. Once a general training exemption exists, the negotiating position of an individual writer becomes much weaker. If permission is legally unnecessary, why would a company pay for it? If permission remains necessary, licensing becomes commercially valuable.

This is not an argument against using AI

I think writers need to keep making an important distinction here. AI-authored fiction and an author using AI as part of a wider creative business are not the same thing.

I write my books. The characters, worlds, story decisions and final creative responsibility are mine. But I can also use AI to help research a subject, visualise a fictional world, organise information, explore marketing ideas, create supporting imagery or handle some of the repetitive work around publishing and promotion.

Those uses do not automatically make the machine the author. They are tools around the creative process. Asking an AI system to generate the novel itself is a different proposition and so is a technology company taking a finished human novel and using it to improve a commercial model.

Using AI around a book is not the same transaction as giving an AI company the right to use the book.

The argument is moving towards permission and payment

That shift is already visible elsewhere in publishing. Publishers’ Licensing Services has argued that the AI debate is becoming a market and that practical licensing solutions now need to follow the principles writers and publishers have spent years debating.

This is where I think the conversation becomes more useful. The question does not have to be whether every author loves or hates AI. It can be much simpler: if a company wants to use a copyrighted book to build or improve a commercial product, what are the terms?

A workable system should be able to answer five things: which works were used, who controlled the relevant rights, what use was permitted, what the creator was paid, and whether the creator had a genuine opportunity to refuse.

That also requires publishers and authors to understand their contracts. Recent disputes over payments from the Anthropic copyright settlement have shown how messy rights records can become when substantial money suddenly attaches to AI use. Rights that once seemed theoretical can become very real very quickly.

The bigger principle

AI companies understandably want access to enormous quantities of high-quality human material. Books are particularly valuable because they contain sustained argument, narrative structure, specialist knowledge, voice and carefully edited language. That value is precisely why authors should not be embarrassed about insisting that their rights matter.

Protecting copyright does not prevent AI innovation. It forces the AI economy to account for one of its inputs. We already accept that businesses pay for computing power, engineers, electricity, data centres and software. Human creative work should not become the one valuable ingredient expected to arrive for nothing.

Australia's decision is therefore worth watching well beyond Australia. If governments hold the line on permission while licensing markets mature, authors may eventually gain something more useful than another court battle: a functioning market in which their work has recognised value.

The question for writers

If an AI company wanted to train on your novel and offered clear disclosure, defined use, meaningful payment and a genuine right to say no, would you license it? Or is there a line you would not cross regardless of the deal?

---

About Rob Frankson

Rob Frankson is a science-fiction author and creator of the Near Galaxy Saga and HMSS Broadside. AI & The Author follows the practical impact of artificial intelligence on writers, publishing, creativity and the business surrounding books while keeping a clear distinction between human authorship and AI used as a supporting tool.

Keep the conversation going

For more AI & The Author briefings, HMSS Broadside, the Near Galaxy Saga and behind-the-scenes work from 121 Minutes, visit the site and subscribe for new posts.

Visit 121 Minutes & Subscribe

AI & Editorial Transparency

AI & The Author is edited and published by Rob Frankson. Artificial intelligence is used to assist with news research, initial drafting, content organisation and supporting imagery. All articles are reviewed and, where necessary, edited by Rob Frankson before publication. The opinions, editorial position and final decision to publish remain the author's.