Zvezdelina Stankova, a math professor at the University of California, Berkeley, admitted to using artificial intelligence to help edit her op-ed in the San Francisco Standard, which criticized studen…
Bradley Emi, CTO of AI text detector Pangram, argues that large language models like ChatGPT, Claude, and Gemini could write as diversely as humans but post-training guardrails cause 'mode collapse,' …
UC Berkeley professor Zvezdelina Stankova admitted to using artificial intelligence to help edit her viral op-ed criticizing the university's admission standards, according to The Daily Californian. T…
A new analysis of 490,000 webpages from Common Crawl snapshots between January 2021 and July 2026 found that 35% of pages with post-ChatGPT publication dates show signs of AI authorship, according to …
University of California, Berkeley, math professor Zvezdelina Stankova admitted to using artificial intelligence to help edit an op-ed for the San Francisco Standard in which she criticized students' …
University of California, Berkeley mathematics professor Zvezdelina Stankova admitted using AI software to help edit her 2,000-word op-ed for the San Francisco Standard that criticized students' math …
Anthropic began adding an imperceptible watermark to text generated by its Claude AI model earlier this month, prompting a surge in demand for tools to remove it. Google searches for "AI watermark rem…
Anthropic announced that text created with Claude, its large language model, will now contain an invisible watermark, making it detectable by A.I. detectors at the individual word level. The change, m…
AI-generated writing remains poor three years after widespread adoption of large language models, according to a blogger's analysis that found ChatGPT's imitation of Scott Alexander's style added 33 w…
AI detectors, including watermarking methods from AI labs and third-party tools like Pangram, are a bad idea because they cannot measure the human effort behind a text, leading to the dismissal of leg…
Elizabeth Tai's blog post 'Substack writers, you need a website!' reached No.1 on Hacker News, a milestone she attributes to prior promotion by WordPress.com's Freshly Pressed page. Tai, a technical c…
AI labs can embed invisible watermarks in generated text by using a secret key to steer word choices, enabling later detection of AI-authored content. The watermark is only detectable with the key, an…
Harvard professor Ricardo Hausmann is facing criticism after the Financial Times added an editor's note to his August 13 op-ed stating that AI was used to condense a longer draft of the column prior t…
Anthropic announced that future Claude models will watermark generated text to comply with the EU AI Act, effective August 2, and the method will not affect output quality, add hidden characters, or i…
Harvard Kennedy School professor Ricardo Hausmann faced backlash after an anti-Trump op-ed he wrote for the Financial Times was found to be 71% AI-generated, according to AI detector Pangram. The FT a…
Pangram, a $20-per-month web tool, correctly identified AI-generated text in dozens of tests by a technology journalist, including flagging AI sentences inside human essays, but it and Hive Detect fai…
Anthropic published a blog post Friday detailing how its Claude chatbot's new text watermarking will work, using the SynthID-Text approach from Google DeepMind and planning a detection API, to comply …
On August 2, 2026, Anthropic launched Claude models that embed a statistical watermark in their output to comply with the EU AI Act's Article 50(2), but the company's own documentation admits the mark…
A new arXiv study (2608.11256v1) finds that commercial AI detectors cannot reliably distinguish AI-assisted editing from full AI drafts, flagging guideline-compliant 'refine abstract only' edits at 64…
Google's Gemini AI model is losing market share to OpenAI's ChatGPT and Anthropic's Claude, according to new market data from Pangram, Similarweb, and OpenRouter. Pangram reports Gemini's share droppe…