By skyforbes 
OpenAI says it aims to radically transform the internet experience with a new web browser called ChatGPT Atlas. Sam Altman describes it as a “once-in-a-decade opportunity”. Atlas puts the chat interface from ChatGPT at the center of the browser, understanding and summarizing the pages you browse and adding an “agent” layer that can take action on your behalf. While this vision sounds powerful, it raises new and unique questions about security and privacy.
What is Atlas? How does it work?
Atlas redefines classic scanning with two key features:
- Agent Mode: ChatGPT can act semi-autonomously within the browser, opening pages, filling out forms, making reservations, editing email texts. For example, if you say “Find a cocktail bar near me and book it for 20:00”, it will explore the options and try to make a reservation.
- Browser Memories: Personalizes responses by recording the sites you visit and your interaction. You can view, manage, turn off or delete these records in the settings. Atlas also has an incognito window, which is not so incognito.
OpenAI says that content is not included in the model training “by default”, but navigation content is processed for personalization. Atlas runs on a Chromium foundation, positioning the chat experience…