正文 · AI 翻译
Thanks to IT Home reader Coje_He for the tip!
IT Home reported on October 17 that OpenAI updated the ChatGPT for Chrome extension on October 15 (version 1.26.1005.11759). ChatGPT can now read multiple tabs open in the Chrome browser and use the content of those pages to answer questions, organize information, or complete tasks across pages. The feature is available to all users, whether or not they subscribe to a paid service.
After a user invokes the browser in a chat, ChatGPT can use currently open webpages as context, for example to compare multiple product pages, compile information from several sources, or analyze work documents.
Citing a blog post, IT Home said the update also introduces more powerful “action” capabilities. In Work mode, ChatGPT is no longer limited to reading webpages: with authorization, it can control tabs to help users fill out forms, send messages, download files, and organize and group tabs, among other tasks.
Because the extension runs in the user's usual browser profile, it can use existing website login sessions to assist directly with workflows in Gmail, LinkedIn, Salesforce, or internal company tools.
However, the convenience comes with broader permission requirements. When installing or enabling these capabilities, Chrome may ask users to let the extension read and modify website data, access page debugging tools, manage tabs and downloads, and even read bookmarks or browsing history or display notifications.
OpenAI says these permissions are necessary for ChatGPT to operate the browser, but users can still control the scope of its access through per-site authorization, allowlists, blocklists, and other measures.
By default, ChatGPT asks for confirmation before accessing each new website. Users can choose “Allow only once,” “Always allow for this website,” “Always allow for all websites,” or deny access. OpenAI labels “Always allow for all websites” as a higher-risk option; browsing history and file uploads and downloads also require separate authorization.