ChatGPT Atlas: How OpenAI’s AI Browser Takes the Web Further

The race to weave AI into mainstream software is accelerating as tech giants shift focus from standalone bots to the core applications people rely on daily.

OpenAI has introduced Atlas, a web browser that seamlessly integrates conversational AI within its interface.

The rollout begins on macOS for Free, Plus, Pro and Go subscribers, with support for Windows, iOS and Android set to follow.

Instead of requiring users to switch tabs or rely on copy-paste, Atlas puts ChatGPT at the heart of the browsing experience.

The company believes “AI gives us a rare moment to rethink what it means to use the web”.

What ChatGPT Atlas can do

The browser leverages ChatGPT’s memory capabilities, drawing on past interactions to shape future responses.

It also debuts what OpenAI refers to as “browser memories” – a feature that captures and retains context from the websites you visit.

OpenAI says the feature enables users to make requests such as: “Find all the job postings I was exploring last week and summarise the key industry trends to help me prepare for interviews.”

For Yogya Kalra, a college student who tested Atlas ahead of launch, the integration is transforming how she studies: “During lectures, I like using practice questions and real-world examples to really understand the material.

“I used to switch between my slides and ChatGPT, taking screenshots just to ask a question. Now ChatGPT instantly understands what I’m looking at, helping me improve my knowledge checks as I go.”

Why agent mode takes the wheel for paying subscribers

The highlight of the release is Agent Mode, allowing ChatGPT to open tabs and navigate websites on its own.

Currently in preview for Plus, Pro and Business users, OpenAI says it’s “now better at researching and analysing, automating tasks, and planning events or booking appointments as you browse”.

Hosting a dinner party? The company illustrates: “You can give the recipe to ChatGPT and ask it to find a grocery store, add all the ingredients to a cart and order them to your house.”

At work, OpenAI says “you can ask ChatGPT to open and read through past team documents, perform new competitive research – and compile insights into a team brief”.

OpenAI is keeping tight control over the agent’s capabilities.

It cannot execute code within browsers, download files or install extensions, nor can it access other applications or local file systems.

When encountering sensitive sites, such as financial platforms, the agent pauses to ensure the user can monitor its activity.

The company acknowledges the security risks, noting that “agents are susceptible to hidden malicious instructions, which may be hidden in places such as a webpage or email with the intention that the instructions override ChatGPT agent’s intended behaviour”.

Such exploits could result in data exposure or unintended actions.

Despite conducting thousands of hours of security testing, OpenAI concedes that “our safeguards will not stop every attack that emerges as AI agents grow in popularity”.

How privacy controls give users the final say

Browser memories are an optional feature.

Users can review or archive them at any time within settings and OpenAI confirms that “deleting browsing history deletes any associated browser memories.”

By default, the company does not use browsing content to train its models, though users can choose to opt in.

A toggle in the address bar gives users control over which sites ChatGPT can access.

Parental controls already set within ChatGPT accounts extend to Atlas, now complemented by additional browser-specific options.

Looking ahead, OpenAI’s roadmap includes support for multiple profiles, enhanced developer tools and new opportunities for Apps SDK developers to increase their presence within the browser.

“This launch marks a step toward a future where most web use happens through agentic systems – where you can delegate the routine and stay focused on what matters most,” the company states.