Abstract
Using the controversy surrounding the Commonwealth Short Story Prize as a starting point, this blog explores how generative AI challenges traditional copyright concepts such as originality, fair use and authorship. Focusing on creative writing, it examines broader questions about what we value when we call something creative, when AI stops being a tool and becomes a crutch, and how the law should respond to these increasingly blurred boundaries.
Introduction
The Commonwealth Short Story Prize had over 7000 applicants. It’s a golden opportunity for unpublished authors from across the world to be recognised for their work. The organisation evidently prides itself on diversity – writers from all corners of the world participate and share their stories, providing a mix of characters diversifying the narrative of what it means to be human. So naturally, it came under fire for a winning story that had the typical hallmarks of being generated by artificial intelligence. Pangram, a detector claiming only 0.2% margin of error, marked it as “fully AI-generated” with “high confidence”. But the criticism went beyond the detector’s verdict. Many readers sensed something strikingly inauthentic in the prose. Notably, it reads like a grocery list of details about Trinidadian culture, and jumps from metaphor to metaphor, a caricature of depth over any real meaning.
Marking the Territory
Yet the Commonwealth Foundation did not act on the allegations. It refused to use AI generators, citing concerns about “consent and artistic ownership.” No matter how accurate detectors claim to be, the risk of a false positive outweighs the benefits. Despite detectors like Turnitin and Pangram stating they don’t store data, sharing privately submitted work with third-party detectors is risky and unethical.
In the world of intellectual property, this is uncharted territory. Who owns the copyright for AI-generated writing? The human who generated the work? The extent of human involvement varies considerably, but some input is necessary. If they provided a skeleton of the plot, or wrote three-fourths of a story and questioned AI about possible endings, it can be argued that the work is original enough to warrant protection.
But what if someone just names a genre and asks for a story? Or provides a vague one-line prompt but gives AI the reins completely from there? Who gets credit there? Can AI claim co-ownership?But these ideas don’t materialise from thin air. Large Language Models need inordinate amounts of data to train, mostly obtained through web scraping (the automated extraction of data from various websites.) Given the non-consensual nature of this extraction, it’s fair to call it pirated material, but big tech companies defend it under the fair use doctrine, an exception meant to allow commercial use of copyrighted work when promoting freedom of expression, like parody or film criticism.
The logical next step is to grant licensing rights to the authors of the scraped data. There are two issues with that: 1) AI abstracts the data into mathematical tokens. It doesn’t permanently hold the original data in its memory. It’s like when a line from a book subconsciously sticks with you, and you use it in another essay. 2) Ideas aren’t copyrightable. If they were, narrative structures, cheesy romcom tropes, character traits, or even the morals in creative work, would be liable.
The courts currently handle cases at the intersection of generative AI and copyright by emphasising human authorship. In Thaler v Perlmutter, The United States Supreme Court declined to review a case asking for copyright protection in AI-generated visual art. It is held by courts as indisputable that the term “author”, meaning the originator of the work, cannot include AI.
But visual art and written work are different mediums by nature. For visual art, the end result is as important as the techniques used to create it. It isn’t only about the abstract seed of an idea in your head, but having the physical skills to bring it to fruition. AI is not a digital art tool like Procreate, where it makes art more convenient, by combining tools like brushes of different sizes and letting you undo when you make a mistake. AI art is more like describing your vision to someone more talented and commissioning them to do your work.
Creative writing is different. The value of the skill is ideation and how it’s conveyed. Whether you write in a notebook, laptop or 17th century typewriter, the result is the same. If you dictate a story to someone and they thoughtlessly type it down, you are still considered the author. There’s no parallel to the element of doing natural to visual mediums. Consequently, whether something is “completely original” is more of a grey area. What about writing tools like Grammarly, or using ChatGPT to refine your work? They completely rewrite some sentences, improve the vocabulary of some others, and extensively polish your work. Here, AI is a tool.
One thing is certain, the ship to dismiss all AI generated work as unoriginal has long sailed. Like it or not, it’s making its way into every step of the creative process, from the brainstorming to the output. To minimise copyright infringement cases, we need to tackle the problem at its source. One approach – accepting using copyrighted works to train AI falls under fair use, and eliminating copyright cases entirely. AI “learns” about existing work so that it can “think” about outputs. It doesn’t completely regurgitate the same material it was trained on, just takes inspiration from it. Supporters argue that it usually makes substantial changes to the original material, and synthesizes from different sources, not blatantly ripping off one. These are two parameters of fair use satisfied. However, this doesn’t negate that unlike critics writing reviews, or clips of news being used to set context, AI can seriously damage the market of the copyrighted material. If AI is capable of producing literature that people are willing to read, publishers replacing human writers can feel like an inevitability.
Another approach – mandating disclosure of what data was used and how it was sourced from LLM developers , shifting towards agreements where copyright authors license their work and are compensated for it, also have their cons, particularly for foundation model providers. It slows down the exponential progress AI has been making and burdens them with additional cost. Whether that’s a bad thing is up to you. But from a perspective of commercial interest, these organizations have no incentive to be regulated. And with the economic growth developments in AI can bring, public administrative bodies have no incentive to regulate either. The risk of regulatory capture makes the possibility of following through with such measures doubtful.
Conclusion
The road we go down is dependent on whether we perceive AI as a leap of mankind, something that will inevitably take over and improve every domain of economy, a net good with acceptable cons. Or, the other shoe drops, and it becomes a threat to life and what makes life worth living, and we have no choice but to exert control. That is yet to be seen. What remains certain is that AI doesn’t just raise issues about copyright, it shakes the foundation of the concept of creativity.
The case-by-case approach, where we look at each instance of how AI uses copyright and determine whether it falls under fair use, can work now. But as AI becomes more capable of mimicking human voices, and starts producing meaningful work, and the line between our voice and what it’s saying becomes increasingly blurred, we need a definitive answer – When does using AI for creative work go from assistance to replacement?
References
The Guardian, ‘’Obvious markers of AI’: doubts raised over winner of short story prize’ (19 May 2026), [https://www.theguardian.com/books/2026/may/19/commonwealth-short-story-prize-winner-doubts-ai-artificial-intelligence]
TechPolicy.Press, ‘How the Emerging Market for AI Training Data is Eroding Big Tech’s ‘Fair Use’ Copyright Defense’ (4 Mar 2025), [https://www.techpolicy.press/how-the-emerging-market-for-ai-training-data-is-eroding-big-techs-fair-use-copyright-defense/]
Jones Walker, ‘OpenAI Loses Privacy Gambit: 20 Million ChatGPT Logs Likely Headed to Copyright Plaintiffs’ (6 Jan 2026), [https://www.joneswalker.com/en/insights/blogs/ai-law-blog/openai-loses-privacy-gambit-20-million-chatgpt-logs-likely-headed-to-copyright-p.html?id=102lzo9]
Ropes & Gray LLP, ‘Web Scraping in the Age of AI: Guidance for Data Owners and Scrapers’ (28 May 2026), [https://www.ropesgray.com/en/insights/alerts/2026/05/web-scraping-in-the-age-of-ai-guidance-for-data-owners-and-scrapers]
Vidhi Centre for Legal Policy, ‘AI and Copyright:’ (22 Jan 2025), [https://vidhilegalpolicy.in/blog/ai-and-copyright/]
Reuters, ‘U.S. Supreme Court declines to hear dispute over copyrights for AI-generated material’ (2 Mar 2026), [https://www.reuters.com/legal/government/us-supreme-court-declines-hear-dispute-over-copyrights-ai-generated-material-2026-03-02/]
Khurana & Khurana, ‘Ownership of AI generated content: A Deep Dive into Copyright Law in India’ (15 Jun 2026), [https://www.khuranaandkhurana.com/ownership-of-ai-generated-content-a-deep-dive-into-copyright-law-in-india]
The Guardian, ‘More than half of UK novelists believe AI will replace their work’ (20 Nov 2025), [https://www.theguardian.com/books/2025/nov/20/more-than-half-of-uk-novelists-believe-ai-will-replace-their-work]
Amlan Mohanty & Shatakratu Sahu, ‘‘India’s Advance on AI Regulation’ (21 Nov 2024), [https://carnegieendowment.org/research/2024/11/indias-advance-on-ai-regulation]