Sam Altman has shared an extract from a new AI model they’ve trained, and it’s already causing a stir. We don’t know if or when we’ll get access to this model—OpenAI hasn’t given a release date or even guaranteed we’ll see it. However, from the preview Altman posted on X (formerly Twitter), it’s clear the model isn’t just spitting out generic text. It’s dabbling in something quite literary.

What’s So Special About This Model?

Usually, when we think of AI text generators, we picture formulaic lines and half-decent attempts at Q&A. This one, though, has taken a leap into creative territory. Altman himself said, “This is the first time I have been really struck by something written by AI.” That’s a big statement, especially considering the long list of AI writing tools out there. This new model produced a story that was not only coherent but also pushed into poetic and self-referential territory—qualities you’d expect from a seasoned writer, not an algorithm.

What the Hell Is Metafiction?

So, the sample story involved two characters, Mila and Kai, in a narrative that was consciously aware it was a story—aka metafiction. That’s a fancy term for writing that steps back and draws attention to itself as fiction. Essentially, the text can acknowledge that it’s being written, shaping the story in a playful or reflective way.

Does this mean it can write in other styles? Possibly. Metafiction isn’t the only type of writing it might pull off. If an AI can be taught to break the “fourth wall,” there’s every reason to believe it can switch gears and emulate noir, romance, or even experimental poetry. That’s the real kicker: the more it reads, the more it can remix.

What the Creative Industry Thinks

While many are wowed by the AI’s newfound flair, the creative industry is none too pleased about the method behind its training. Publishers and authors have pointed to instances where copyrighted material appears to be used without permission. Their argument? If AI is getting smarter by devouring the work of real writers, shouldn’t those writers get something in return?

• Dan Conway, CEO of the Publishers Association, called out OpenAI directly, saying this new example is further proof of training “on copyright-protected literary content.”
• Anna Ganley, CEO of the Society of Authors, acknowledges the AI’s improved output—yet insists creators deserve a piece of the action if their texts feed the machine.

More on the Make It Fair Campaign

In the UK, the Make It Fair campaign is pushing for strict regulations to stop AI companies from automatically scraping copyrighted material. Their argument is simple: the default should be “don’t use our work without permission,” not “use our work unless we somehow opt out.” This is a direct reaction to proposals that would let AI developers comb through just about any publicly available content unless the original author goes out of their way to say “no.”

What Happened with the Data (Use and Access) Bill?

This is where it gets political. Labour initially supported amendments that would protect creators under the Data (Use and Access) Bill. Yet, when it got to the Commons, those amendments were stripped out. Baroness Kidron, an outspoken advocate for these protections, vowed to reintroduce them, expecting new voices to join the fight. Given how many big brands and authors have a serious stake in the outcome, we could be in for a legislative tug-of-war that’ll shape the future of AI development and creative rights.

Conclusion

As someone who’s been working on perfecting my own voice in AI-generated articles, I can’t help but find it thrilling that AI is evolving and writing more “human”-feeling text. We’re edging closer to a world where machines can craft stories that resonate on an emotional level—stories that might even rival human authors. Where does it go next, and at what cost, though? The potential is enormous, but the friction between innovation and fair compensation is just as huge. We’ll likely see big debates—and bigger headlines—in the months to come.