In its ongoing competition with Google, OpenAI has unveiled its latest offering—an AI-enhanced web browser known as ChatGPT Atlas. This new tool was introduced during a livestream demonstration, following an earlier teaser featuring an enigmatic video of browser tabs against a white background.
Starting today, ChatGPT Atlas is accessible globally for users on macOS, while versions for Windows, iOS, and Android are expected to follow soon. However, the browser's "agent mode," a special feature, is currently exclusive to ChatGPT Plus and Pro subscribers. OpenAI CEO Sam Altman expressed hope that this new approach would revolutionize how people use the internet, envisioning a chat-based experience within a web browser as analogous to future internet usage.
The livestream also featured several OpenAI staff members, including Will Ellsworth, who focuses on post-training research; Adam Fry, the product lead for ChatGPT Search; Ben Goodger, who has previously contributed to developing Google Chrome and Mozilla Firefox; Ryan O'Rouke, an interface designer; Justin Rushing, formerly of Apple; and Pranav Vishnu.
Fry highlighted key attributes of the browser, such as its memory feature, which personalizes the browsing experience, and the agent mode that enables ChatGPT to perform actions for users like booking reservations, arranging flights, or editing documents. Users can oversee and control their “memories” via the browser settings and open incognito windows for private browsing.
The agent mode is an evolution of OpenAI’s earlier endeavors with AI assistance, such as the Operator tool, an initial version that allowed ChatGPT to operate a computer on a user's behalf, and its successor, ChatGPT Agent, which was engineered to handle complex tasks, including shopping and more, though not always flawlessly.
When users click a link from a search result in Atlas, the browser defaults to a dual-screen display showing both the webpage and a ChatGPT transcript, providing a companion experience. However, users can choose to disable this feature if preferred. During the presentation, OpenAI staff also showcased features like page summarization and "cursor chat," which refines sentences in emails through an intuitive button click.
"This browser's performance is outstanding—it's swift, seamless, and superbly user-friendly," Altman remarked enthusiastically.
The competition among AI-enabled browsers has intensified, with OpenAI previously announcing its SearchGPT search engine prototype in July 2024. Recently, AI-enhanced browsers are regarded as the latest trend in AI technology.
Last summer, Perplexity introduced its Comet browser, an AI-powered alternative designed to streamline web browsing and task execution. Instead of traditional search results, users receive a simplified "answer engine" that offers relevant website links and responses to inquiries. Comet is also capable of scanning open tabs, summarizing videos, organizing email inboxes, and even making purchases on Amazon.
In September, Google shared plans to further integrate its Gemini AI assistant within Chrome, enabling it to handle "tedious tasks," like shopping or scheduling, though a launch date remains unspecified.