An internal version of the Astra model has made significant strides in mathematics and computer science, while new safety rules for ChatGPT have been introduced in response to pressure from the literary community over copyright infringement.
On Sunday, August 2, 2026, OpenAI announced two significant moves. According to a post on Hacker News, an internal version of their next major model, named Astra, has successfully solved as many as 10 major open problems in the fields of mathematics, quantum complexity, and theoretical computer science. At the same time, as reported by Kurir.rs, the popular chatbot ChatGPT has been given new safety rules that prohibit it from mimicking the writing style of living authors, putting an end to what had been a controversial practice.
Although not yet officially unveiled to the public, the Astra model is already demonstrating remarkable capabilities in scientific reasoning. Solving 10 major open problems represents a huge step forward for artificial intelligence in the domain of complex thinking. "We believe this will be a major leap for scientific reasoning," reads an OpenAI statement shared on Hacker News.
In parallel with this technological achievement, OpenAI has also introduced significant changes to how ChatGPT operates. For years, it was enough to type a command like "write me a story in the style of Stephen King or George R. R. Martin" for the algorithm to generate, in a second, text that strikingly resembles the style of famous writers. According to Kurir, that practice has now come to an end.
This decision is a direct consequence of immense pressure from the author community and copyright infringement lawsuits. Famous writers, including "Game of Thrones" creator George R. R. Martin and John Grisham, have accused AI companies of illegally using millions of pages of their copyrighted books to train language models. They argued that artificial intelligence thus directly parasitizes their years of work and creates unfair competition.
The new safety rules embedded in ChatGPT work by having the algorithm now recognize requests for direct imitation of specific authors. If a user requests text written in the recognizable style of a living writer, the system will either refuse the request or offer a more generic version that does not copy original phrases and sentence rhythms. Engineers have implemented complex filters to distinguish general genre styles, such as epic fantasy or crime fiction, from the authentic voice of individual authors.
It is important to note that this ban does not mean the end of creative writing. Users can still ask ChatGPT to write a dark thriller, a science fiction story, or a lyrical poem, but without referencing specific names of contemporary creators. Literary classics that have entered the public domain can still serve as inspiration.
OpenAI's decision has elicited mixed opinions. Writers' associations worldwide have welcomed this move as an important victory in the fight for intellectual property protection. On the other hand, some AI enthusiasts and researchers believe that such restrictions hamper the full potential of language models, arguing that style imitation has always been a fundamental way humans learn to write.
