Microsoft markets its Copilot AI assistant as a productivity tool for consumers and businesses alike. However, the product’s terms of use tell a different story. According to these terms, which were last updated in October 2025, “Copilot is for entertainment purposes only. It can make mistakes, and it may not work as intended. Don’t rely …
Google’s AI Overviews are accurate about 91 percent of the time. This sounds good at first, but Tripp Mickle and colleagues report for The New York Times that this still means the search engine delivers tens of millions of incorrect answers every hour. The New York Times commissioned AI startup Oumi to test Google’s system …
A new study finds that artificial intelligence systems are excessively agreeable when users seek advice on personal and interpersonal matters. Myra Cheng reports for Stanford University that large language models (LLMs) consistently side with users, even when their behavior is harmful or illegal. The researchers tested 11 major LLMs, including ChatGPT, Claude, Gemini, and DeepSeek. …
Wikipedia has updated its guidelines to prohibit editors from using artificial intelligence to write or rewrite article content. The English-language version of the site introduced the change after months of editors struggling with a rise in AI-generated articles. The new policy stems from a finding that text produced by large language models (LLMs) frequently breaks …
The AI image marketplace Tess.Design paid artists a 50% royalty each time their style was used to generate an image. Julia Enthoven writes how the platform ran for 20 months before closing in January 2026. Tess.Design allowed artists to submit their work to fine-tune an AI model. That model was then listed on a public …
About 10,000 authors have published a book with no content to protest against AI companies using their work without permission. Dan Milmo reports for The Guardian that contributors include Nobel laureate Kazuo Ishiguro, Philippa Gregory and Richard Osman. The only content in “Don’t Steal This Book” is a list of the contributors’ names. Copies are …
Grammarly, a subscription-based writing assistant owned by the company Superhuman, has disabled an AI feature called “expert review” after widespread criticism from writers, academics, and journalists whose names and likenesses were used without their consent. The feature, which launched in August, presented users with AI-generated writing suggestions attributed to real people, including living authors like …
A new study warns that intensive use of AI tools is causing a distinct form of mental exhaustion among workers, separate from traditional burnout. Researchers at Boston Consulting Group report findings from a survey of 1,488 full-time U.S.-based workers across industries, roles, and seniority levels. The study identifies a phenomenon the researchers call “AI brain …
Public skepticism toward artificial intelligence is deepening across the United States, even as tech companies pour billions of dollars into the technology. Protests, lawsuits, political campaigns, and union contracts are emerging as tools for people pushing back against an industry that many Americans feel is moving too fast and ignoring their concerns. A 2025 Pew …
A new term is emerging in the debate around artificial intelligence and writing: semantic ablation. Claudio Nastruzzi writes for The Register that AI tools do not just add errors to text. They also systematically destroy what makes writing distinctive in the first place. Semantic ablation describes how AI models erode high-value, precise, or unconventional language …