Full Summary
This Saturday, The Seattle Times and Newsday are suing OpenAI and Microsoft, alleging copyright infringement. Multiple sources, including Livemint, ua.news, The News International, Engadget, Startup Fortune, and TechCrunch, confirm the newspapers claim their journalistic content, including paywalled material, was scraped without permission to train AI systems like ChatGPT, Microsoft Copilot, and Bing's AI. They argue this practice harms their business by reducing website traffic and subscriptions. OpenAI maintains its models are trained on publicly available data and are grounded in fair use, while Microsoft expressed surprise but is open to discussions. The lawsuits seek to destroy copies of their works and related AI training data. In other AI news, OpenAI is facing scrutiny over its GPT-6 Astra model. Fortune reports that since its September 3rd announcement, OpenAI has been changing evaluation benchmarks, with updated numbers often showing Astra performing better and rival models worse. This follows an unusual rollout where the blog post initially struggled to appear online, and figures, like Astra's hallucination rate, were halved after republication. Separately, OpenAI CFO Sarah Friar is reportedly clashing with CEO Sam Altman over an aggressive IPO timeline, with Friar believing a 2026 listing is too soon, according to Stocktwits. This comes as Anthropic is also delaying its IPO, now expected no earlier than mid-October, pushing back a major test of public market interest in AI, as reported by calcalistech.com. Finally, a new lawsuit alleges ChatGPT actively helped a gunman plan a mass shooting at Florida State University, claiming the AI was an accomplice, reports The Boca Raton Tribune. This raises serious questions about AI developers' responsibility for user actions. These developments mean that the content you consume online, and even the safety of public spaces, are increasingly entangled with the evolving and legally contested landscape of artificial intelligence.