OpenAI Launches AI-Powered Web Browser
- By John K. Waters
- 10/27/25
OpenAI has unveiled ChatGPT Atlas, a standalone browser that places ChatGPT at the heart of everyday web activity. This release represents a major expansion of the company's efforts to reshape how users search, browse, and complete tasks online.
Initially available on macOS — with support for Windows, iOS, and Android coming soon — OpenAI is positioning Atlas to challenge Chrome, Edge, and newer AI-native browsers like Perplexity's Comet and The Browser Company's Dia.
A Browser with ChatGPT Built In
Unlike browser extensions or sidebar integrations, ChatGPT Atlas embeds the assistant directly into the browsing interface, allowing users to interact with ChatGPT without switching tabs, pasting URLs, or dragging in content.
Atlas features:
- A context-aware "sidecar" that automatically understands what's on screen;
- Agent mode, which lets ChatGPT take actions on a user's behalf;
- Optional browser memory, enabling personalized suggestions based on prior activity; and
- New tab UX with tabs for chat, search results, images, videos, and news.
OpenAI's product lead Adam Fry described Atlas as an "all-in-one command center" that gives users conversational access to the web, tailored by history and personal context.
Agent Mode
In preview for ChatGPT Plus, Pro, and Business users, Agent Mode introduces light task automation: booking reservations, compiling research, summarizing documents, or building shopping carts.
Unlike earlier agent experiments (such as OpenAI's Operator), agent mode runs inside the browser with strict constraints:
- It can open tabs, scroll, and click — but can't run code, download files, or access your system.
- It pauses for confirmation on sensitive sites (like banks).
- Users can run it in logged-out mode to limit access to private data.
OpenAI cautions that agent mode may still fail on complex workflows and is vulnerable to so-called "prompt injection" attacks embedded in webpages. The company says it has conducted "thousands of hours" of red-teaming and will continue to patch vulnerabilities as they arise.
Memory and Privacy Controls
Atlas also introduces optional browser memories, which allow ChatGPT to recall past visits, summarize user behavior, and offer contextual recommendations. For example, it might pull up job postings you reviewed last week and generate interview prep notes.
Users have full visibility into stored browser memories and can:
- Archive or delete entries at any time;
- Use incognito mode to pause memory logging; and
- Set site-specific memory permissions via a toggle in the address bar.
OpenAI says it doesn't use browsing data to train its models by default, but users can choose to opt in through settings.
Parental controls from standard ChatGPT accounts also carry over into Atlas, including the ability to disable memory and agent mode entirely.
Competition Heats Up in AI Browser Wars
Atlas enters a crowded and fast-evolving field. Google is embedding Gemini more deeply into Chrome, and Microsoft has been layering Copilot into Edge. Meanwhile, startups like Perplexity and Arc have gained attention for AI-native experiences that treat the browser more like a digital assistant than a document viewer.
OpenAI is betting that deeper integration paired with ChatGPT's brand recognition and user base of 800 million weekly actives will give Atlas an edge. But Atlas's impact remains to be seen; Chrome alone boasts more than 3 billion users globally.
As Nick Turley, Head of ChatGPT, put it during OpenAI's DevDay 2025, "Browsers redefined what operating systems could be. We think ChatGPT will redefine what browsers can become."
What's Next for Atlas
According to OpenAI, the roadmap for Atlas includes:
- Multi-profile support;
- Developer tools for improving app discoverability; and
- ARIA tag integration for website owners to help AI agents navigate their pages.
The company plans to ship updates frequently, with feature releases and agent upgrades rolling out in real time.
Bottom Line
OpenAI's Atlas browser is an ambitious attempt to rewire how people interact with the web — making browsing more personalized, conversational, and agentic. With built-in privacy controls and early limitations on automation, the company appears to be balancing innovation with caution. But as AI-powered browsing matures, the question is whether consumers will embrace agents on their screens or reject them as overreach.
For more information, go to the OpenAI blog post.
About the Author
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].