OpenAI's ChatGPT now refuses direct requests to copy specific authors' writing styles. The shift addresses potential copyright concerns while the AI still generates content with "broad qualities" of particular writers.
ChatGPT has implemented new restrictions preventing users from explicitly asking it to replicate an author's distinct style. The move comes amid growing legal scrutiny over generative AI's use of copyrighted works.
Users can no longer request outputs written "in the style of" named authors. However, the system still accommodates requests for writing with generalized characteristics—such as "academic tone" or "journalistic brevity"—that don't directly target specific creators.
The distinction reflects OpenAI's attempt to navigate copyright liability. Courts have yet to definitively rule on whether style emulation constitutes infringement, but the change signals the company's caution as litigation over AI training data expands.
Authors including Sarah Silverman and others have filed lawsuits claiming unauthorized use of their work to train large language models. OpenAI's new guardrails suggest the company is refining its approach to minimize legal exposure while preserving core functionality.
TIME magazine is detecting AI crawlers and serving them a different version of its website that includes advertisements. The practice raises questions about how publishers are monetizing bot traffic.
The American Federation of Teachers is partnering with major tech companies to fund AI literacy training for educators, addressing growing concerns about artificial intelligence in classrooms.
Hark has previewed a new browser use agent designed to automate web tasks. The company claims its solution outperforms competitors on both speed and cost.
Open-source models now outperform OpenAI's frontier GPT-5.6 Sol on retrieval tasks while costing a fraction of the price. Neon and Castform demonstrated the efficiency gap in a new benchmark.