OpenAI has launched ChatGPT Atlas, a new web browser that tightly integrates ChatGPT into the browsing experience. Available here today for macOS users on Free, Plus, Pro, and Go plans, Atlas effectively turns the web itself into a workspace where ChatGPT can assist in real time. The company calls it “the browser with ChatGPT built in.”
Atlas offers a new tab page where users can either ask ChatGPT questions or enter a URL, blurring the line between search and chat. It adds tabs for traditional search links, images, videos, and news results, all presented within a single interface. A sidebar allows ChatGPT to summarize or analyze content on any page without leaving the site. Users can even call up ChatGPT in text boxes or forms to write or edit directly inside a web app.
One of Atlas’s more ambitious features is browser memories, which let ChatGPT remember what you’ve browsed to offer contextual suggestions or revisit old pages. These memories are stored privately under the user’s account, with full controls for viewing, archiving, or deleting them. Parents can disable these memories or Agent mode entirely through the new parental controls section.
Agent mode, available in preview for Plus, Pro, and Business users, allows ChatGPT to perform multi-step tasks inside the browser, such as planning a trip, researching topics, or even adding items to a shopping cart. OpenAI stresses that user oversight remains central, since the system must ask permission before acting, and it cannot access files, extensions, or saved passwords. It also includes a “logged out” mode to prevent access to personal accounts or cookies.
Privacy and safety remain front and center. Atlas does not use browsing data to train models by default, and even when training is enabled, users can control which sites ChatGPT sees. OpenAI warns, however, that agentic systems carry risks such as malicious web instructions and urges users to monitor activity when using agent mode.
Personally, I’m excited to try ChatGPT Atlas. I’ve covered a lot of AI companies over the years, and OpenAI has consistently shown a stronger commitment to transparency and user control than most of its competitors. I trust it with my data and privacy far more than I would a random AI browser from another developer. That trust makes the difference between curiosity and actual adoption.
That said, it’s disappointing that Windows users have to wait, and Linux isn’t even mentioned at all. Linux users are usually among the earliest to embrace new technology, and ignoring them feels like a missed opportunity. OpenAI says versions for Windows, iOS, and Android are coming soon, but so far, there’s no word on desktop Linux support.
I don’t get companies and people you talk about Linux they want to bit at you. Mac is not that far ahead of Linux users and now atlas makes a browser for mac lol. I must say one thing about Linux and Zorin they might be slow when it comes to updates to the kernel but they know how to build a simple and beautiful operating system that is easy to navigate. I prefer Zorin 18 over Windows and the gap in software compatibility is shrinking fast. You can find more 95% of software that is on Windows for Linux now and even better you can find a few and open source software that will do the same job.