Full Summary
This Saturday morning, OpenAI is rolling out a new "Computer History" feature for its ChatGPT desktop app on macOS, allowing the AI to remember your activity across applications and websites. Both Dainik Jagran and Gizbot confirm this optional tool aims to provide ChatGPT with more context, making it easier to recover past work or continue tasks without repeated explanations. This feature, available for Pro, Business, and Enterprise users, records interaction events like clicks and typing, but notably, it does not continuously capture screenshots or audio, a key difference from Microsoft's Recall. Multiple sources, including Notebookcheck and ETV Bharat, highlight that this data is stored locally on your Mac, and users can exclude specific apps, pause tracking, or delete their history. However, Notebookcheck also points out that while the raw event files are processed on OpenAI's servers, they are not encrypted by Computer History, raising privacy concerns. In other AI news, innovation-village.com reports that Claude is demonstrating a massive advantage in minimizing AI hallucinations, with its flagship Fable 5 model showing 61% accuracy compared to ChatGPT's GPT 5.6 Sol at 59%. This suggests Claude could be a more reliable tool for factual accuracy. Meanwhile, ChatGPT is also rolling out new features for paid users, including Google Drive support, the ability to create quizzes on any topic, and a reservation search function, as detailed by digit.in. But here's the thing: while AI tools advance, so do threats. Kaspersky warns of over 45,000 fake ChatGPT cyberattacks exploiting user trust, with cybercriminals spreading malware through fake AI applications. This means your digital interactions, whether with new AI features or seemingly legitimate downloads, require heightened vigilance to protect your personal data and device security.