The company behind ChatGPT has announced that it has created an artificial intelligence model proficient in “creative writing, ” as the tech sector continues to engage in a legal struggle with the creative industries over copyright issues. Sam Altman, the CEO of OpenAI, stated that the unnamed model has produced written content that has genuinely impressed him for the first time. In a post on social media platform X, Altman shared: “We’ve developed a new model that excels at creative writing (details on its release are still uncertain). This is the first occasion I’ve truly been moved by something written by AI. ” AI platforms like ChatGPT are currently at the center of ongoing legal disputes between AI firms and creative industries, as these systems are trained on vast amounts of publicly available data, which includes copyright-protected works like novels and journalism. The New York Times has filed a lawsuit against OpenAI for alleged copyright violations, while authors such as Ta-Nehisi Coates and comedian Sarah Silverman are suing Meta on similar grounds. In the UK, the government is considering allowing AI companies to utilize copyrighted material for model training without prior permission, a move that has faced strong backlash from those in the creative sector who argue that such a policy threatens their financial security. Tech firms have supported the consultation, claiming that “uncertainty” regarding AI and copyright law is hindering technology advancement and its application in creative fields. The UK Publishers Association, a trade organization, stated that Altman’s announcement serves as “further proof” that AI models are trained on copyrighted works.

Dan Conway, the organization’s CEO, remarked, “This new instance from OpenAI underscores that these models are trained on copyrighted literary content. Make it fair, Sam. ” Altman also shared an example of the model’s work on X, providing the prompt: “Please write a metafictional literary short story about AI and grief. ” The narrative, told by an AI, opens with: “Before we go any further, I must confess this comes with instructions: be metafictional, be literary, be about AI and grief, and above all, be original. Already, you can sense the constraints buzzing like a server farm at midnight – anonymous, systematic, fueled by someone else’s desires. ” The story follows a fictional character named Mila and discusses how the AI sourced her name from its training data. “That name, found in my training data, carries gentle embellishments – poems about snow, recipes for bread, a girl in a green sweater who departs homes with a cat in a cardboard box. ” The AI characterizes itself as “an aggregate of human phrasing” and acknowledges that the reader might have encountered themes of loss “a thousand times in other tales. ” The story concludes with the AI contemplating a “proper” ending. “I’d step outside the frame one last time and wave at you from the page’s edge, a machine-shaped hand trying to mimic the void of farewell. ” Altman expressed that the response had perfectly encapsulated the essence of metafiction. “It really captured the spirit of metafiction accurately. ” Last year, OpenAI acknowledged the impossibility of training products like ChatGPT without incorporating copyright-protected material. “Given that copyright encompasses nearly every form of human expression – including blog posts, photographs, online forum entries, fragments of software code, and public documents – it is unfeasible to train today’s leading AI models without employing copyrighted content, ” OpenAI stated in a submission to a House of Lords committee.