OpenAI releases bot-tom feeding browser with ChatGPT built in

Why experience the web for yourself when there's so much privacy to surrender?

In a bid to grab even more eyeballs, OpenAI has finally released Atlas, its long-teased, ChatGPT-powered web browser. Surfing the web may never be the same now that a bot is doing it for you – while training itself at the same time.

Released on Tuesday for macOS, OpenAI's Atlas offers the ability to direct ChatGPT via text or voice to interact with web content and to navigate through websites. (Atlas for Android, iOS, and Windows is expected soon.) In a livestream presentation, OpenAI CEO and AI hypemaster Sam Altman said that Atlas represents an opportunity to "rethink what a browser can be about." In this case, "rethink" means allowing ChatGPT to observe, mediate, and direct browsing sessions.

"With Atlas, ChatGPT can come with you anywhere across the web – helping you in the window right where you are, understanding what you're trying to do, and completing tasks for you, all without copying and pasting or leaving the page," OpenAI enthused in a blog post. "Your ChatGPT memory is built in, so conversations can draw on past chats and details to help you get new things done."

As described in the release notes, Atlas makes ChatGPT available in a sidebar to analyze or summarize webpage content. ChatGPT can also be invoked from within form fields loaded in an Atlas-rendered page. Users have the option to enable browser memories – past browsing history – and to deny ChatGPT access to specific sites.

Plus, Pro, and Business users have access to agent mode, which enables Atlas to perform more complicated multi-step functions. 

"In agent mode, ChatGPT can complete end to end tasks for you like researching a meal plan, making a list of ingredients, and adding the groceries to a shopping cart ready for delivery," the release notes explain, adding that ChatGPT will ask before taking important actions and can be interrupted. This is a reasonable attempt to prevent the AI model from accidentally purchasing two tons of creamed corn and making other unexpected decisions.

As with rival Perplexity's Comet browser, which debuted in July, OpenAI's hope is that web users will delegate their interactions with websites to AI agents that act on their behalf, even making purchases. But OpenAI also wants to ensure that ChatGPT, the big behemoth of bots, has access to the largest possible audience. 

Browsers represent the primary consumer access point for the web, and browser makers like Google and Microsoft have been integrating favored AI services to promote usage. Google has built Gemini into Chrome while Microsoft has integrated Copilot with Edge.

Smaller browser companies like Brave and Opera have taken similar steps, while Mozilla is offering a choice of AI models in Firefox. Only Vivaldi has refrained from bot bombing its customers.

It's not obvious that Atlas will do much to change AI market share. As noted in VC firm Menlo Ventures' State of Consumer AI Report 2025, "About a quarter of Americans report using ChatGPT (28 percent) or Google Gemini (23 percent) in the past six months, indicating that consumer adoption today is largely driven by embedded products with existing distribution."

It's hard to get people to change their habits.

Owning a browser may also prove useful for OpenAI if it chooses to enter the online ad business – a scenario that Altman previously dismissed and now looks more likely. With OpenAI losing money at such a rapid rate, expect that ad revenue will get a serious look.

As a browser, Atlas has a few quirks. It offers Developer Tools via the View menu, but they can't be opened on a New Tab page – presumably because the default ChatGPT input box is not structured as a web page. And if you try to get rid of the suggested prompts that load in a New Tab window, the suggestions reappear upon reload.

Atlas is based on Google's Chromium open source project and it does support Chrome extensions. Unlike Google's own Chrome browser, which (for now) offers a way to disable AI Mode via chrome://flags, there's no atlas://flags settings page to turn off ChatGPT. But then that may be the point.

OpenAI very much wants people to set Atlas as their default browser and has gone so far as to offer increased ChatGPT usage limits for those who keep Atlas set as the default for seven consecutive calendar days.

All your browsing data are belong to us

The release notes include a section titled "Privacy and data controls," which is noteworthy, because at launch there's no Privacy menu in the Atlas settings.

The release notes also claim that "By default, we don't use the content you browse to train our models." 

That was not this reporter's experience – upon installing and opening Atlas for the first time, the menu Data Controls -> Model improvement defaulted to "On."

Screenshot of Atlas defaulting to sharing data for AI training

Screenshot of Atlas defaulting to sharing data for AI training - Click to enlarge

"On" means "Allow your content to be used to train our models, which makes ChatGPT better for everyone who uses it. We take steps to protect your privacy. Learn more."

It's perhaps worth noting that the statement "We take steps to protect your privacy" rivals "Security, we've heard of it" and "Your call is very important to us" as vague, non-binding commitments.

Still, OpenAI acknowledges that there are some potential concerns.

"Despite all of the power and awesome capabilities that you get with sharing your browser with ChatGPT, that also poses an entirely new set of risks," said Pranav Vishnu, product lead for ChatGPT Atlas, during the video presentation.

Vishnu said that's why Atlas comes with limitations like only being able to operate on your tabs and not being able to execute code on your computer or access local files.

Simon Willison, a widely-followed open source developer, opined in a blog post that he continues to find browser agents deeply confusing.

"The security and privacy risks involved here still feel insurmountably high to me – I certainly won't be trusting any of these products until a bunch of security researchers have given them a very thorough beating," he wrote, adding that he also finds AI browsers "pretty unexciting to use." ®

More about

TIP US OFF

Send us news


Other stories you might like