AI and Publishing
The publishing industry in Australia and across the world has faced many threats to its production. When e-books emerged, there was an industry-wide scramble to standardise them, as well as to determine if physical print would be impacted. Then, with the introduction of audiobooks, there was a similar fear for the industry’s future. Within the traditional publishing industry, each technological invention brings uncertainty. How will print media change? Will there still be a demand? How can we integrate this technology into publishing? Since the rapid development of generative artificial intelligence (AI) and large language models (LLMs), there has been an upheaval as the industry attempts to adapt to these technological advancements.
LLMs are a category of deep learning models, which work as statistical prediction machines that constantly predict the following word in a sentence based upon patterns from data that it has been fed through training. The process of training requires immense amounts of data, which is collected from websites, books and other forms of media. Data collection is one of the largest issues attributed to LLMs, especially within the publishing industry. Since LLMs require data to operate—taking information and determining patterns to later make predictions when required—the ethics behind this data collection is fraught.
The public accessibility to generative AI models has already brought publishing standards into question.
In 2023, tech conglomerate Meta held an internal meeting to discuss the potential acquisition of Simon & Schuster. Aimed at accessing vast quantities of copywritten data, Meta’s focus was on training their own text-generating AI models.
In 2026, manuscript Call Me, I’ll Hide the Body by Jerry Falade was withdrawn from production due to its agents being unable to authenticate its process.
In 2026, Hachette pulled Shy Girl by Mia Ballard from US release after accusations of AI use.
The prevalence of AI has resulted in major publishing houses like Penguin Random House, Macmillian and Sourcebooks recruiting AI engineers.
Harvey notes that the future of trade publishing is intimately entwined with AI and that within an Australian context, the increasingly aggressive development of AI presents a threat to the trade publishing industry. It is suggested that for publishers to engage effectively with AI infrastructure, there needs to be an increase in transparency and democratic procedures.
The Australian Publishers Association (APA) says it ‘embraces the possibilities presented by AI technology’ while acknowledging it presents ‘serious legal and ethical challenges’, listing four principles it would like the government to legislate to protect the Australian publishing industry:
Policies must be underpinned by a clearly defined ethical framework
Transparency is key
Ensure appropriate incentives and protections for creators and rights holders
Policy settings must be balanced
Without an overarching framework, companies’ use and integration of AI remains unmonitored. This has led to different interpretations of ethical AI use. Unlike Amazon, which modified their own policies to repurpose users’ data for AI training, Meta was too impatient to negotiate intellectual property law with trade publishers, instead feeding pirated content into their algorithms.
Involving AI in Australia’s trade publishing industry undermines its integrity. Through computerising human processes that engage with culture and knowledge as well as reducing products to consumable data, AI companies challenge the frameworks that made trade publishing successful. Even if AI technology was integrated into established publishing systems, framing Meta’s technology as a neutral, universal entity also denies meaningful external input into how such AI models are engaged with and developed.
With the specifics of how and when LLMs will be integrated into the industry, the future of publishing sits at a crossroads. Considering the versatility and adaptability of generative AI, from an industry perspective, even if an individual intends to reject it, their industry has likely already adopted it into their processes. LLMs—and the data centres that train them—require significant amounts of water, threaten the protections of copyrighted content and undermine the integrity of the Australian trade publishing industry. As a direct response to generative AI, people are embracing old tech, valuing human-made products and appreciating creativity. Within an industry which is entirely shaped by the limitless imagination of people, there is hope publishing will centre humanness as the only legitimate force shaping stories and art.